# Table of Contents

- [Welcome to RunAnywhere - RunAnywhere Documentation](#welcome-to-runanywhere-runanywhere-documentation)
- [Best Practices - RunAnywhere Documentation](#best-practices-runanywhere-documentation)
- [Configuration - RunAnywhere Documentation](#configuration-runanywhere-documentation)
- [Installation - RunAnywhere Documentation](#installation-runanywhere-documentation)
- [Introduction - RunAnywhere Documentation](#introduction-runanywhere-documentation)
- [chat() - RunAnywhere Documentation](#chat-runanywhere-documentation)
- [Streaming TTS - RunAnywhere Documentation](#streaming-tts-runanywhere-documentation)
- [generate() - RunAnywhere Documentation](#generate-runanywhere-documentation)
- [Streaming STT - RunAnywhere Documentation](#streaming-stt-runanywhere-documentation)
- [Configuration - RunAnywhere Documentation](#configuration-runanywhere-documentation)
- [generateStream() - RunAnywhere Documentation](#generatestream-runanywhere-documentation)
- [Voices - RunAnywhere Documentation](#voices-runanywhere-documentation)
- [Error Handling - RunAnywhere Documentation](#error-handling-runanywhere-documentation)
- [Voice Activity Detection - RunAnywhere Documentation](#voice-activity-detection-runanywhere-documentation)
- [STT Options - RunAnywhere Documentation](#stt-options-runanywhere-documentation)
- [Installation - RunAnywhere Documentation](#installation-runanywhere-documentation)
- [System Prompts - RunAnywhere Documentation](#system-prompts-runanywhere-documentation)
- [synthesize() - RunAnywhere Documentation](#synthesize-runanywhere-documentation)
- [Quick Start - RunAnywhere Documentation](#quick-start-runanywhere-documentation)
- [Introduction - RunAnywhere Documentation](#introduction-runanywhere-documentation)
- [transcribe() - RunAnywhere Documentation](#transcribe-runanywhere-documentation)
- [Error Handling - RunAnywhere Documentation](#error-handling-runanywhere-documentation)
- [chat() - RunAnywhere Documentation](#chat-runanywhere-documentation)
- [generate() - RunAnywhere Documentation](#generate-runanywhere-documentation)
- [Introduction - RunAnywhere Documentation](#introduction-runanywhere-documentation)
- [Installation - RunAnywhere Documentation](#installation-runanywhere-documentation)
- [Best Practices - RunAnywhere Documentation](#best-practices-runanywhere-documentation)
- [Streaming TTS - RunAnywhere Documentation](#streaming-tts-runanywhere-documentation)
- [SDK Overview - RunAnywhere Documentation](#sdk-overview-runanywhere-documentation)
- [System Prompts - RunAnywhere Documentation](#system-prompts-runanywhere-documentation)
- [Streaming STT - RunAnywhere Documentation](#streaming-stt-runanywhere-documentation)
- [Voice Activity Detection - RunAnywhere Documentation](#voice-activity-detection-runanywhere-documentation)
- [transcribe() - RunAnywhere Documentation](#transcribe-runanywhere-documentation)
- [generateStream() - RunAnywhere Documentation](#generatestream-runanywhere-documentation)
- [Voices - RunAnywhere Documentation](#voices-runanywhere-documentation)
- [STT Options - RunAnywhere Documentation](#stt-options-runanywhere-documentation)
- [Quick Start - RunAnywhere Documentation](#quick-start-runanywhere-documentation)
- [Installation - RunAnywhere Documentation](#installation-runanywhere-documentation)
- [Introduction - RunAnywhere Documentation](#introduction-runanywhere-documentation)
- [synthesize() - RunAnywhere Documentation](#synthesize-runanywhere-documentation)
- [Voice Agent - RunAnywhere Documentation](#voice-agent-runanywhere-documentation)
- [chat() - RunAnywhere Documentation](#chat-runanywhere-documentation)
- [Quick Start - RunAnywhere Documentation](#quick-start-runanywhere-documentation)
- [Voice Agent - RunAnywhere Documentation](#voice-agent-runanywhere-documentation)
- [Configuration - RunAnywhere Documentation](#configuration-runanywhere-documentation)
- [chat() - RunAnywhere Documentation](#chat-runanywhere-documentation)
- [System Prompts - RunAnywhere Documentation](#system-prompts-runanywhere-documentation)
- [Quick Start - RunAnywhere Documentation](#quick-start-runanywhere-documentation)
- [Best Practices - RunAnywhere Documentation](#best-practices-runanywhere-documentation)
- [Error Handling - RunAnywhere Documentation](#error-handling-runanywhere-documentation)
- [STT Options - RunAnywhere Documentation](#stt-options-runanywhere-documentation)
- [Error Handling - RunAnywhere Documentation](#error-handling-runanywhere-documentation)
- [Best Practices - RunAnywhere Documentation](#best-practices-runanywhere-documentation)
- [generate() - RunAnywhere Documentation](#generate-runanywhere-documentation)
- [generate() - RunAnywhere Documentation](#generate-runanywhere-documentation)
- [Quick Start - RunAnywhere Documentation](#quick-start-runanywhere-documentation)
- [System Prompts - RunAnywhere Documentation](#system-prompts-runanywhere-documentation)
- [Voices - RunAnywhere Documentation](#voices-runanywhere-documentation)
- [transcribe() - RunAnywhere Documentation](#transcribe-runanywhere-documentation)
- [Voices - RunAnywhere Documentation](#voices-runanywhere-documentation)
- [Voice Activity Detection - RunAnywhere Documentation](#voice-activity-detection-runanywhere-documentation)
- [STT Options - RunAnywhere Documentation](#stt-options-runanywhere-documentation)
- [Streaming STT - RunAnywhere Documentation](#streaming-stt-runanywhere-documentation)
- [transcribe() - RunAnywhere Documentation](#transcribe-runanywhere-documentation)
- [Streaming TTS - RunAnywhere Documentation](#streaming-tts-runanywhere-documentation)
- [Configuration - RunAnywhere Documentation](#configuration-runanywhere-documentation)
- [synthesize() - RunAnywhere Documentation](#synthesize-runanywhere-documentation)
- [generateStream() - RunAnywhere Documentation](#generatestream-runanywhere-documentation)
- [Streaming STT - RunAnywhere Documentation](#streaming-stt-runanywhere-documentation)
- [synthesize() - RunAnywhere Documentation](#synthesize-runanywhere-documentation)
- [generateStream() - RunAnywhere Documentation](#generatestream-runanywhere-documentation)
- [Voice Activity Detection - RunAnywhere Documentation](#voice-activity-detection-runanywhere-documentation)
- [Streaming TTS - RunAnywhere Documentation](#streaming-tts-runanywhere-documentation)
- [Voice Agent - RunAnywhere Documentation](#voice-agent-runanywhere-documentation)
- [Voice Agent - RunAnywhere Documentation](#voice-agent-runanywhere-documentation)

---

# Welcome to RunAnywhere - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Getting Started

Welcome to RunAnywhere

On this page

*   [What is RunAnywhere?](https://docs.runanywhere.ai/#what-is-runanywhere)
    
*   [Core AI Capabilities](https://docs.runanywhere.ai/#core-ai-capabilities)
    
*   [Voice Agent Pipeline](https://docs.runanywhere.ai/#voice-agent-pipeline)
    
*   [Why RunAnywhere?](https://docs.runanywhere.ai/#why-runanywhere)
    
*   [Supported Platforms](https://docs.runanywhere.ai/#supported-platforms)
    
*   [Get Started](https://docs.runanywhere.ai/#get-started)
    

![RunAnywhere Documentation](https://mintcdn.com/runanywhere/PgeUY8BfGFmDDdg8/images/intro-light.jpg?w=2500&fit=max&auto=format&n=PgeUY8BfGFmDDdg8&q=85&s=3985d4e1665d9c39c1d5db2d3be9b312) ![RunAnywhere Documentation](https://mintcdn.com/runanywhere/PgeUY8BfGFmDDdg8/images/intro-dark.jpg?w=2500&fit=max&auto=format&n=PgeUY8BfGFmDDdg8&q=85&s=228e2b777a634dbf98510877eee10300)

[​](https://docs.runanywhere.ai/#what-is-runanywhere)

What is RunAnywhere?
-----------------------------------------------------------------------------

**RunAnywhere** is a production-grade, on-device AI platform that enables developers to run AI models directly on mobile devices. Our SDKs provide a unified API for running AI inference locally — ensuring minimal latency, maximum privacy, and offline capability.

All AI inference runs **100% on-device**. Once models are downloaded, no network connection is required for inference.

[​](https://docs.runanywhere.ai/#core-ai-capabilities)

Core AI Capabilities
------------------------------------------------------------------------------

Every RunAnywhere SDK provides access to these powerful AI features:

[LLM (Large Language Model)\
--------------------------\
\
On-device text generation with streaming support, structured outputs, and system prompts](https://docs.runanywhere.ai/swift/llm/chat)
[STT (Speech-to-Text)\
--------------------\
\
Real-time audio transcription with multiple backends and multi-language support](https://docs.runanywhere.ai/swift/stt/transcribe)
[TTS (Text-to-Speech)\
--------------------\
\
Neural and system voice synthesis with customizable voice, pitch, rate, and volume](https://docs.runanywhere.ai/swift/tts/synthesize)
[VAD (Voice Activity Detection)\
------------------------------\
\
Real-time speech detection for hands-free interactions](https://docs.runanywhere.ai/swift/vad)

### 

[​](https://docs.runanywhere.ai/#voice-agent-pipeline)

Voice Agent Pipeline

Build complete voice-powered experiences with our integrated **Voice Agent** that orchestrates: This enables full voice conversation flows with streaming and batch processing modes.

[​](https://docs.runanywhere.ai/#why-runanywhere)

Why RunAnywhere?
---------------------------------------------------------------------

Privacy by Design

Audio and text data never leaves the device unless you explicitly configure it. Only anonymous analytics are collected by default. Your users’ data stays on their device.

Low Latency

On-device inference eliminates network round-trips. Get AI responses in milliseconds, not seconds. Perfect for real-time voice interactions and responsive UIs.

Offline Capable

Once models are downloaded, your app works completely offline. No internet required for inference. Ideal for apps used in areas with poor connectivity.

Plugin Architecture

Backend engines are optional modules—include only what you need. Keep your app binary size minimal by importing only the capabilities you use.

[​](https://docs.runanywhere.ai/#supported-platforms)

Supported Platforms
----------------------------------------------------------------------------

RunAnywhere provides native SDKs for all major mobile platforms:

[![Swift](https://mintcdn.com/runanywhere/PgeUY8BfGFmDDdg8/images/code-languages/swift/swift_color.svg?w=2500&fit=max&auto=format&n=PgeUY8BfGFmDDdg8&q=85&s=71f71dff148590897979dba4207b8ca1)Swift SDK\
\
**iOS, macOS, tvOS, watchOS** Native Swift API with Metal acceleration for Apple Silicon](https://docs.runanywhere.ai/swift/introduction)
[![Kotlin](https://mintcdn.com/runanywhere/PgeUY8BfGFmDDdg8/images/code-languages/kotlin/kotlin_color.svg?w=2500&fit=max&auto=format&n=PgeUY8BfGFmDDdg8&q=85&s=4121f6a5e21b7378aced8ffea82962c2)Kotlin SDK\
\
**Android** Native Kotlin API with GPU acceleration support](https://docs.runanywhere.ai/kotlin/introduction)
[![React Native](https://upload.wikimedia.org/wikipedia/commons/a/a7/React-icon.svg)React Native\
\
**iOS + Android** TypeScript-first with full type safety for cross-platform apps](https://docs.runanywhere.ai/react-native/introduction)
[![Flutter](https://mintcdn.com/runanywhere/PgeUY8BfGFmDDdg8/images/code-languages/flutter/flutter_color.svg?w=2500&fit=max&auto=format&n=PgeUY8BfGFmDDdg8&q=85&s=93c53557df38a96ececaa120569159e0)Flutter SDK\
\
**iOS + Android** Dart SDK for cross-platform Flutter applications](https://docs.runanywhere.ai/flutter/introduction)

[​](https://docs.runanywhere.ai/#get-started)

Get Started
------------------------------------------------------------

1

[](https://docs.runanywhere.ai/#)

Choose Your SDK

Select the SDK that matches your tech stack from the options above

2

[](https://docs.runanywhere.ai/#)

Install the SDK

Follow the installation guide for your chosen platform

3

[](https://docs.runanywhere.ai/#)

Initialize & Build

Initialize the SDK and start building AI-powered features

[Quick Start Guide\
-----------------\
\
Get up and running in 5 minutes](https://docs.runanywhere.ai/quickstart)
[SDK Comparison\
--------------\
\
Compare features across all platforms](https://docs.runanywhere.ai/sdks)

[Quick Start](https://docs.runanywhere.ai/quickstart)

⌘I

---

# Best Practices - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/best-practices#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Flutter SDK

Best Practices

On this page

*   [Model Selection](https://docs.runanywhere.ai/flutter/best-practices#model-selection)
    
*   [Memory Management](https://docs.runanywhere.ai/flutter/best-practices#memory-management)
    
*   [Unload Unused Models](https://docs.runanywhere.ai/flutter/best-practices#unload-unused-models)
    
*   [Check Available Memory](https://docs.runanywhere.ai/flutter/best-practices#check-available-memory)
    
*   [Clean Up Old Models](https://docs.runanywhere.ai/flutter/best-practices#clean-up-old-models)
    
*   [Streaming for Better UX](https://docs.runanywhere.ai/flutter/best-practices#streaming-for-better-ux)
    
*   [Preload Models](https://docs.runanywhere.ai/flutter/best-practices#preload-models)
    
*   [Error Handling](https://docs.runanywhere.ai/flutter/best-practices#error-handling)
    
*   [Test on Physical Devices](https://docs.runanywhere.ai/flutter/best-practices#test-on-physical-devices)
    
*   [Download Management](https://docs.runanywhere.ai/flutter/best-practices#download-management)
    
*   [Show Progress](https://docs.runanywhere.ai/flutter/best-practices#show-progress)
    
*   [Download on WiFi](https://docs.runanywhere.ai/flutter/best-practices#download-on-wifi)
    
*   [Voice Assistant Tips](https://docs.runanywhere.ai/flutter/best-practices#voice-assistant-tips)
    
*   [Optimize Latency](https://docs.runanywhere.ai/flutter/best-practices#optimize-latency)
    
*   [Handle Background/Foreground](https://docs.runanywhere.ai/flutter/best-practices#handle-background%2Fforeground)
    
*   [Performance Monitoring](https://docs.runanywhere.ai/flutter/best-practices#performance-monitoring)
    
*   [Summary Checklist](https://docs.runanywhere.ai/flutter/best-practices#summary-checklist)
    
*   [See Also](https://docs.runanywhere.ai/flutter/best-practices#see-also)
    

Follow these best practices to build high-quality AI-powered Flutter apps.

**Reference Implementation:** The [Flutter Starter Example](https://github.com/RunanywhereAI/flutter-starter-example)
 demonstrates all these best practices in a production-ready app.

[​](https://docs.runanywhere.ai/flutter/best-practices#model-selection)

Model Selection
------------------------------------------------------------------------------------------

Choose the right model size for your use case:

| Model Size | RAM Required | Use Case | Example |
| --- | --- | --- | --- |
| 360M–500M (Q8) | ~500MB | Fast chat, commands | SmolLM2-360M |
| 1B–3B (Q4/Q6) | 1–2GB | Balanced quality/speed | Qwen2.5-1.5B |
| 7B (Q4) | 4–5GB | High quality | Llama-3.2-7B |

**Start small** — Use the smallest model that meets your quality needs. You can always upgrade later.

[​](https://docs.runanywhere.ai/flutter/best-practices#memory-management)

Memory Management
----------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/flutter/best-practices#unload-unused-models)

Unload Unused Models

Copy

Ask AI

    // Unload when not in use
    await RunAnywhere.unloadModel();
    await RunAnywhere.unloadSTTModel();
    await RunAnywhere.unloadTTSVoice();
    

### 

[​](https://docs.runanywhere.ai/flutter/best-practices#check-available-memory)

Check Available Memory

Copy

Ask AI

    final storageInfo = await RunAnywhere.getStorageInfo();
    final freeSpace = storageInfo.deviceStorage.freeSpace;
    
    // Ensure enough space before downloading
    final model = await RunAnywhere.availableModels()
        .then((m) => m.firstWhere((m) => m.id == 'my-model'));
    
    if (model.downloadSize != null && model.downloadSize! * 2 > freeSpace) {
      showError('Not enough storage. Need ${model.downloadSize! * 2} bytes.');
    }
    

### 

[​](https://docs.runanywhere.ai/flutter/best-practices#clean-up-old-models)

Clean Up Old Models

Copy

Ask AI

    // Delete unused models
    await RunAnywhere.deleteStoredModel('old-model-id');
    
    // Get storage usage
    final models = await RunAnywhere.getDownloadedModelsWithInfo();
    for (final model in models) {
      print('${model.id}: ${(model.size / 1024 / 1024).toStringAsFixed(1)} MB');
    }
    

[​](https://docs.runanywhere.ai/flutter/best-practices#streaming-for-better-ux)

Streaming for Better UX
----------------------------------------------------------------------------------------------------------

Always prefer streaming for chat interfaces:

Copy

Ask AI

    // ✅ Good: Streaming shows progress
    final result = await RunAnywhere.generateStream(prompt);
    await for (final token in result.stream) {
      setState(() => _response += token);
    }
    
    // ❌ Avoid: User waits with no feedback
    final response = await RunAnywhere.generate(prompt);
    setState(() => _response = response.text);
    

[​](https://docs.runanywhere.ai/flutter/best-practices#preload-models)

Preload Models
----------------------------------------------------------------------------------------

Load models during idle time, not when user needs them:

Copy

Ask AI

    class MyApp extends StatefulWidget {
      @override
      _MyAppState createState() => _MyAppState();
    }
    
    class _MyAppState extends State<MyApp> {
      @override
      void initState() {
        super.initState();
        _preloadModels();  // Load in background
      }
    
      Future<void> _preloadModels() async {
        // Download if needed
        final models = await RunAnywhere.availableModels();
        final llm = models.firstWhere((m) => m.id == 'smollm2-360m-q8_0');
    
        if (!llm.isDownloaded) {
          await for (final p in RunAnywhere.downloadModel(llm.id)) {
            if (p.state.isCompleted) break;
          }
        }
    
        // Preload into memory
        await RunAnywhere.loadModel(llm.id);
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/best-practices#error-handling)

Error Handling
----------------------------------------------------------------------------------------

Always handle errors gracefully:

Copy

Ask AI

    Future<void> generateResponse(String prompt) async {
      try {
        final result = await RunAnywhere.generate(prompt);
        setState(() => _response = result.text);
      } on SDKError catch (e) {
        setState(() => _error = _getUserFriendlyMessage(e));
      } catch (e) {
        setState(() => _error = 'Something went wrong. Please try again.');
      }
    }
    
    String _getUserFriendlyMessage(SDKError error) {
      // Convert technical errors to user-friendly messages
      switch (error.code) {
        case SDKErrorCode.componentNotReady:
          return 'AI is loading. Please wait...';
        case SDKErrorCode.generationFailed:
          return 'Could not generate response. Try again.';
        default:
          return 'An error occurred.';
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/best-practices#test-on-physical-devices)

Test on Physical Devices
------------------------------------------------------------------------------------------------------------

**Always test on real devices** — Emulators are significantly slower for AI inference. Performance you see in emulators is not representative of real-world usage.

*   iOS Simulator: 5-10x slower than device
*   Android Emulator: 3-5x slower than device

[​](https://docs.runanywhere.ai/flutter/best-practices#download-management)

Download Management
--------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/flutter/best-practices#show-progress)

Show Progress

Copy

Ask AI

    Widget _buildDownloadProgress(String modelId) {
      return StreamBuilder<DownloadProgress>(
        stream: RunAnywhere.downloadModel(modelId),
        builder: (context, snapshot) {
          if (!snapshot.hasData) return CircularProgressIndicator();
    
          final progress = snapshot.data!;
          return Column(
            children: [\
              LinearProgressIndicator(value: progress.percentage),\
              Text('${(progress.percentage * 100).toStringAsFixed(0)}%'),\
              Text('${(progress.bytesDownloaded / 1024 / 1024).toStringAsFixed(1)} MB'),\
            ],
          );
        },
      );
    }
    

### 

[​](https://docs.runanywhere.ai/flutter/best-practices#download-on-wifi)

Download on WiFi

Copy

Ask AI

    import 'package:connectivity_plus/connectivity_plus.dart';
    
    Future<bool> shouldDownload() async {
      final connectivity = await Connectivity().checkConnectivity();
      return connectivity == ConnectivityResult.wifi;
    }
    
    Future<void> downloadWithCheck(String modelId) async {
      if (!await shouldDownload()) {
        final proceed = await showDialog<bool>(
          context: context,
          builder: (ctx) => AlertDialog(
            title: Text('Download on Cellular?'),
            content: Text('This model is large. Download on WiFi to save data.'),
            actions: [\
              TextButton(onPressed: () => Navigator.pop(ctx, false), child: Text('Wait for WiFi')),\
              TextButton(onPressed: () => Navigator.pop(ctx, true), child: Text('Download Now')),\
            ],
          ),
        );
        if (proceed != true) return;
      }
    
      await for (final p in RunAnywhere.downloadModel(modelId)) {
        if (p.state.isCompleted) break;
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/best-practices#voice-assistant-tips)

Voice Assistant Tips
----------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/flutter/best-practices#optimize-latency)

Optimize Latency

Copy

Ask AI

    // Load all components during initialization
    await Future.wait([\
      RunAnywhere.loadSTTModel('whisper-tiny-en'),\
      RunAnywhere.loadModel('smollm2-360m-q8_0'),\
      RunAnywhere.loadTTSVoice('piper-en-us-amy'),\
    ]);
    

### 

[​](https://docs.runanywhere.ai/flutter/best-practices#handle-background/foreground)

Handle Background/Foreground

Copy

Ask AI

    class VoiceApp extends StatefulWidget {
      @override
      _VoiceAppState createState() => _VoiceAppState();
    }
    
    class _VoiceAppState extends State<VoiceApp> with WidgetsBindingObserver {
      VoiceSessionHandle? _session;
    
      @override
      void initState() {
        super.initState();
        WidgetsBinding.instance.addObserver(this);
      }
    
      @override
      void didChangeAppLifecycleState(AppLifecycleState state) {
        if (state == AppLifecycleState.paused) {
          _session?.stop();  // Stop when app goes to background
        }
      }
    
      @override
      void dispose() {
        WidgetsBinding.instance.removeObserver(this);
        _session?.stop();
        super.dispose();
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/best-practices#performance-monitoring)

Performance Monitoring
--------------------------------------------------------------------------------------------------------

Track performance metrics:

Copy

Ask AI

    final result = await RunAnywhere.generate(prompt);
    
    // Log performance
    analytics.logEvent('generation', {
      'tokens': result.tokensUsed,
      'latency_ms': result.latencyMs,
      'tokens_per_second': result.tokensPerSecond,
      'model': result.modelUsed,
    });
    

[​](https://docs.runanywhere.ai/flutter/best-practices#summary-checklist)

Summary Checklist
----------------------------------------------------------------------------------------------

Before Launch

*   [ ]  Test on physical devices (iOS and Android)
*   [ ]  Handle all error cases with user-friendly messages
*   [ ]  Implement download progress UI
*   [ ]  Add WiFi check for large downloads
*   [ ]  Test with slow network conditions

Performance

*   [ ]  Use smallest adequate model size - \[ \] Preload models during idle time - \[ \] Use streaming for chat interfaces - \[ \] Unload unused models - \[ \] Monitor memory usage

User Experience

*   [ ]  Show loading states clearly
*   [ ]  Provide cancel option for long operations
*   [ ]  Handle background/foreground transitions
*   [ ]  Save conversation history locally
*   [ ]  Allow model selection for power users

[​](https://docs.runanywhere.ai/flutter/best-practices#see-also)

See Also
----------------------------------------------------------------------------

[Configuration\
-------------\
\
SDK configuration](https://docs.runanywhere.ai/flutter/configuration)
[Error Handling\
--------------\
\
Handle errors gracefully](https://docs.runanywhere.ai/flutter/error-handling)

[Error Handling](https://docs.runanywhere.ai/flutter/error-handling)
[Expo Integration Guide](https://docs.runanywhere.ai/vibe-coding/expo-integration)

⌘I

---

# Configuration - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/configuration#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Flutter SDK

Configuration

On this page

*   [SDK Initialization](https://docs.runanywhere.ai/flutter/configuration#sdk-initialization)
    
*   [Environment Modes](https://docs.runanywhere.ai/flutter/configuration#environment-modes)
    
*   [Development Mode (Default)](https://docs.runanywhere.ai/flutter/configuration#development-mode-default)
    
*   [Production Mode](https://docs.runanywhere.ai/flutter/configuration#production-mode)
    
*   [Check SDK State](https://docs.runanywhere.ai/flutter/configuration#check-sdk-state)
    
*   [Generation Options](https://docs.runanywhere.ai/flutter/configuration#generation-options)
    
*   [Generation Parameters](https://docs.runanywhere.ai/flutter/configuration#generation-parameters)
    
*   [Voice Session Configuration](https://docs.runanywhere.ai/flutter/configuration#voice-session-configuration)
    
*   [Model Registration](https://docs.runanywhere.ai/flutter/configuration#model-registration)
    
*   [LLM Models (GGUF)](https://docs.runanywhere.ai/flutter/configuration#llm-models-gguf)
    
*   [STT Models](https://docs.runanywhere.ai/flutter/configuration#stt-models)
    
*   [TTS Voices](https://docs.runanywhere.ai/flutter/configuration#tts-voices)
    
*   [Storage Management](https://docs.runanywhere.ai/flutter/configuration#storage-management)
    
*   [Reset SDK](https://docs.runanywhere.ai/flutter/configuration#reset-sdk)
    
*   [Environment Variables](https://docs.runanywhere.ai/flutter/configuration#environment-variables)
    
*   [See Also](https://docs.runanywhere.ai/flutter/configuration#see-also)
    

Configure the RunAnywhere SDK for different environments and use cases.

[​](https://docs.runanywhere.ai/flutter/configuration#sdk-initialization)

SDK Initialization
-----------------------------------------------------------------------------------------------

Copy

Ask AI

    await RunAnywhere.initialize(
      apiKey: 'your-api-key',                    // Optional for development
      baseURL: 'https://api.runanywhere.ai',     // Optional for development
      environment: SDKEnvironment.production,    // development | staging | production
    );
    

[​](https://docs.runanywhere.ai/flutter/configuration#environment-modes)

Environment Modes
---------------------------------------------------------------------------------------------

| Environment | Description | API Key | Logging | Telemetry |
| --- | --- | --- | --- | --- |
| `.development` | Local development | Optional | Verbose | Disabled |
| `.staging` | Testing with services | Required | Info | Enabled |
| `.production` | Production deployment | Required | Minimal | Enabled |

### 

[​](https://docs.runanywhere.ai/flutter/configuration#development-mode-default)

Development Mode (Default)

Copy

Ask AI

    // No API key needed - perfect for local development
    await RunAnywhere.initialize();
    
    // Or explicitly set development mode
    await RunAnywhere.initialize(
      environment: SDKEnvironment.development,
    );
    

### 

[​](https://docs.runanywhere.ai/flutter/configuration#production-mode)

Production Mode

Copy

Ask AI

    await RunAnywhere.initialize(
      apiKey: 'your-production-api-key',
      baseURL: 'https://api.runanywhere.ai',
      environment: SDKEnvironment.production,
    );
    

[​](https://docs.runanywhere.ai/flutter/configuration#check-sdk-state)

Check SDK State
-----------------------------------------------------------------------------------------

Copy

Ask AI

    // Version
    print('SDK Version: ${RunAnywhere.version}');
    
    // Initialization state
    print('Initialized: ${RunAnywhere.isSDKInitialized}');
    print('Active: ${RunAnywhere.isActive}');
    print('Environment: ${RunAnywhere.environment}');
    
    // Model states
    print('LLM loaded: ${RunAnywhere.isModelLoaded}');
    print('Current model: ${RunAnywhere.currentModelId}');
    print('STT loaded: ${RunAnywhere.isSTTModelLoaded}');
    print('TTS loaded: ${RunAnywhere.isTTSVoiceLoaded}');
    print('Voice Agent ready: ${RunAnywhere.isVoiceAgentReady}');
    

[​](https://docs.runanywhere.ai/flutter/configuration#generation-options)

Generation Options
-----------------------------------------------------------------------------------------------

Configure text generation behavior:

Copy

Ask AI

    final options = LLMGenerationOptions(
      maxTokens: 256,              // Maximum tokens to generate
      temperature: 0.7,            // Randomness (0.0–2.0)
      topP: 0.95,                  // Nucleus sampling
      stopSequences: ['END'],      // Stop at these sequences
      systemPrompt: 'You are a helpful assistant.',
    );
    
    final result = await RunAnywhere.generate(prompt, options: options);
    

### 

[​](https://docs.runanywhere.ai/flutter/configuration#generation-parameters)

Generation Parameters

| Parameter | Type | Default | Range | Description |
| --- | --- | --- | --- | --- |
| `maxTokens` | `int` | 100 | 1–4096 | Maximum output length |
| `temperature` | `double` | 0.8 | 0.0–2.0 | Higher = more creative |
| `topP` | `double` | 1.0 | 0.0–1.0 | Nucleus sampling |
| `stopSequences` | `List<String>` | `[]` | \-  | Stop generation triggers |
| `systemPrompt` | `String?` | `null` | \-  | System context |

[​](https://docs.runanywhere.ai/flutter/configuration#voice-session-configuration)

Voice Session Configuration
-----------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    final config = VoiceSessionConfig(
      silenceDuration: 1.5,     // Seconds of silence before processing
      speechThreshold: 0.03,    // Audio level for speech detection
      autoPlayTTS: true,        // Auto-play synthesized audio
      continuousMode: true,     // Resume listening after TTS
    );
    
    final session = await RunAnywhere.startVoiceSession(config: config);
    

[​](https://docs.runanywhere.ai/flutter/configuration#model-registration)

Model Registration
-----------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/flutter/configuration#llm-models-gguf)

LLM Models (GGUF)

Copy

Ask AI

    LlamaCpp.addModel(
      id: 'my-llm-model',
      name: 'My LLM Model',
      url: 'https://huggingface.co/.../model.gguf',
      memoryRequirement: 500000000,  // 500MB
      supportsThinking: false,        // Chain-of-thought support
    );
    

### 

[​](https://docs.runanywhere.ai/flutter/configuration#stt-models)

STT Models

Copy

Ask AI

    Onnx.addModel(
      id: 'my-stt-model',
      name: 'My STT Model',
      url: 'https://example.com/stt-model.tar.gz',
      modality: ModelCategory.speechRecognition,
      memoryRequirement: 75000000,  // 75MB
    );
    

### 

[​](https://docs.runanywhere.ai/flutter/configuration#tts-voices)

TTS Voices

Copy

Ask AI

    Onnx.addModel(
      id: 'my-tts-voice',
      name: 'My TTS Voice',
      url: 'https://example.com/tts-voice.tar.bz2',
      modality: ModelCategory.textToSpeech,
      memoryRequirement: 50000000,  // 50MB
    );
    

[​](https://docs.runanywhere.ai/flutter/configuration#storage-management)

Storage Management
-----------------------------------------------------------------------------------------------

Copy

Ask AI

    // Get storage information
    final storageInfo = await RunAnywhere.getStorageInfo();
    print('Free space: ${storageInfo.deviceStorage.freeSpace} bytes');
    
    // Get downloaded models with sizes
    final models = await RunAnywhere.getDownloadedModelsWithInfo();
    for (final model in models) {
      print('${model.id}: ${model.size} bytes');
    }
    
    // Delete a model
    await RunAnywhere.deleteStoredModel('old-model-id');
    
    // Refresh model discovery
    await RunAnywhere.refreshDiscoveredModels();
    

[​](https://docs.runanywhere.ai/flutter/configuration#reset-sdk)

Reset SDK
-----------------------------------------------------------------------------

Copy

Ask AI

    // Reset SDK state (for testing or reinitialization)
    RunAnywhere.reset();
    

[​](https://docs.runanywhere.ai/flutter/configuration#environment-variables)

Environment Variables
-----------------------------------------------------------------------------------------------------

For production apps, store sensitive configuration securely:

Copy

Ask AI

    import 'package:flutter_dotenv/flutter_dotenv.dart';
    
    await dotenv.load();
    
    await RunAnywhere.initialize(
      apiKey: dotenv.env['RUNANYWHERE_API_KEY'],
      baseURL: dotenv.env['RUNANYWHERE_BASE_URL'],
      environment: SDKEnvironment.production,
    );
    

[​](https://docs.runanywhere.ai/flutter/configuration#see-also)

See Also
---------------------------------------------------------------------------

[Error Handling\
--------------\
\
Handle SDK errors](https://docs.runanywhere.ai/flutter/error-handling)
[Best Practices\
--------------\
\
Optimization tips](https://docs.runanywhere.ai/flutter/best-practices)

[Voice Agent](https://docs.runanywhere.ai/flutter/voice-agent)
[Error Handling](https://docs.runanywhere.ai/flutter/error-handling)

⌘I

---

# Installation - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/installation#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Flutter SDK

Installation

On this page

*   [Packages on pub.dev](https://docs.runanywhere.ai/flutter/installation#packages-on-pub-dev)
    
*   [Add Dependencies](https://docs.runanywhere.ai/flutter/installation#add-dependencies)
    
*   [iOS Setup (Required)](https://docs.runanywhere.ai/flutter/installation#ios-setup-required)
    
*   [1\. Update Podfile](https://docs.runanywhere.ai/flutter/installation#1-update-podfile)
    
*   [2\. Add Microphone Permission](https://docs.runanywhere.ai/flutter/installation#2-add-microphone-permission)
    
*   [3\. Install Pods](https://docs.runanywhere.ai/flutter/installation#3-install-pods)
    
*   [Android Setup](https://docs.runanywhere.ai/flutter/installation#android-setup)
    
*   [Verify Installation](https://docs.runanywhere.ai/flutter/installation#verify-installation)
    
*   [Starter Example](https://docs.runanywhere.ai/flutter/installation#starter-example)
    
*   [Next Steps](https://docs.runanywhere.ai/flutter/installation#next-steps)
    

[​](https://docs.runanywhere.ai/flutter/installation#packages-on-pub-dev)

Packages on pub.dev
------------------------------------------------------------------------------------------------

The RunAnywhere Flutter SDK consists of three packages:

| Package | Description | pub.dev |
| --- | --- | --- |
| [`runanywhere`](https://pub.dev/packages/runanywhere) | Core SDK with APIs and infrastructure | [![pub package](https://img.shields.io/pub/v/runanywhere.svg)](https://pub.dev/packages/runanywhere) |
| [`runanywhere_llamacpp`](https://pub.dev/packages/runanywhere_llamacpp) | LLM text generation (GGUF models) | [![pub package](https://img.shields.io/pub/v/runanywhere_llamacpp.svg)](https://pub.dev/packages/runanywhere_llamacpp) |
| [`runanywhere_onnx`](https://pub.dev/packages/runanywhere_onnx) | STT, TTS, VAD (ONNX models) | [![pub package](https://img.shields.io/pub/v/runanywhere_onnx.svg)](https://pub.dev/packages/runanywhere_onnx) |

[​](https://docs.runanywhere.ai/flutter/installation#add-dependencies)

Add Dependencies
------------------------------------------------------------------------------------------

Add the packages you need to your `pubspec.yaml`:

*   LLM Only
    
*   STT/TTS Only
    
*   All Features
    

Add these dependencies to `pubspec.yaml`:

Copy

Ask AI

    dependencies:
      runanywhere: ^0.15.11
      runanywhere_llamacpp: ^0.15.11
    

Add these dependencies to `pubspec.yaml`:

Copy

Ask AI

    dependencies:
      runanywhere: ^0.15.11
      runanywhere_onnx: ^0.15.11
    

Add these dependencies to `pubspec.yaml`:

Copy

Ask AI

    dependencies:
      runanywhere: ^0.15.11
      runanywhere_llamacpp: ^0.15.11
      runanywhere_onnx: ^0.15.11
    

Then run:

Copy

Ask AI

    flutter pub get
    

[​](https://docs.runanywhere.ai/flutter/installation#ios-setup-required)

iOS Setup (Required)
------------------------------------------------------------------------------------------------

After adding the packages, update your iOS configuration:

### 

[​](https://docs.runanywhere.ai/flutter/installation#1-update-podfile)

1\. Update Podfile

Open `ios/Podfile` and update:

ios/Podfile

Copy

Ask AI

    # Set minimum iOS version to 14.0
    platform :ios, '14.0'
    
    target 'Runner' do
      # REQUIRED: Add static linkage
      use_frameworks! :linkage => :static
    
      flutter_install_all_ios_pods File.dirname(File.realpath(__FILE__))
    end
    
    post_install do |installer|
      installer.pods_project.targets.each do |target|
        flutter_additional_ios_build_settings(target)
        target.build_configurations.each do |config|
          config.build_settings['IPHONEOS_DEPLOYMENT_TARGET'] = '14.0'
          # Required for microphone permission (STT/Voice features)
          config.build_settings['GCC_PREPROCESSOR_DEFINITIONS'] ||= [\
            '$(inherited)',\
            'PERMISSION_MICROPHONE=1',\
          ]
        end
      end
    end
    

Without `use_frameworks! :linkage => :static`, you will encounter “symbol not found” errors at runtime.

### 

[​](https://docs.runanywhere.ai/flutter/installation#2-add-microphone-permission)

2\. Add Microphone Permission

Add to `ios/Runner/Info.plist` (for STT/Voice features):

ios/Runner/Info.plist

Copy

Ask AI

    <key>NSMicrophoneUsageDescription</key>
    <string>This app needs microphone access for speech recognition</string>
    

### 

[​](https://docs.runanywhere.ai/flutter/installation#3-install-pods)

3\. Install Pods

Copy

Ask AI

    cd ios && pod install && cd ..
    

[​](https://docs.runanywhere.ai/flutter/installation#android-setup)

Android Setup
------------------------------------------------------------------------------------

Add microphone permission to `android/app/src/main/AndroidManifest.xml`:

android/app/src/main/AndroidManifest.xml

Copy

Ask AI

    <uses-permission android:name="android.permission.RECORD_AUDIO" />
    

[​](https://docs.runanywhere.ai/flutter/installation#verify-installation)

Verify Installation
------------------------------------------------------------------------------------------------

Run your app to verify everything is set up correctly:

Copy

Ask AI

    flutter run
    

[​](https://docs.runanywhere.ai/flutter/installation#starter-example)

Starter Example
----------------------------------------------------------------------------------------

Want to see a complete working example? Check out our official Flutter starter app:[Flutter Starter Example\
-----------------------\
\
Clone and run a complete example app with LLM, STT, TTS, and Voice Agent](https://github.com/RunanywhereAI/flutter-starter-example)

Copy

Ask AI

    git clone https://github.com/RunanywhereAI/flutter-starter-example.git
    cd flutter-starter-example
    flutter pub get
    flutter run
    

[​](https://docs.runanywhere.ai/flutter/installation#next-steps)

Next Steps
------------------------------------------------------------------------------

[Quick Start\
-----------\
\
Build your first AI feature →](https://docs.runanywhere.ai/flutter/quick-start)

[Introduction](https://docs.runanywhere.ai/flutter/introduction)
[Quick Start](https://docs.runanywhere.ai/flutter/quick-start)

⌘I

---

# Introduction - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/introduction#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Flutter SDK

Introduction

On this page

*   [Overview](https://docs.runanywhere.ai/flutter/introduction#overview)
    
*   [Key Capabilities](https://docs.runanywhere.ai/flutter/introduction#key-capabilities)
    
*   [Core Philosophy](https://docs.runanywhere.ai/flutter/introduction#core-philosophy)
    
*   [Features](https://docs.runanywhere.ai/flutter/introduction#features)
    
*   [Language Models (LLM)](https://docs.runanywhere.ai/flutter/introduction#language-models-llm)
    
*   [Speech-to-Text (STT)](https://docs.runanywhere.ai/flutter/introduction#speech-to-text-stt)
    
*   [Text-to-Speech (TTS)](https://docs.runanywhere.ai/flutter/introduction#text-to-speech-tts)
    
*   [Voice Activity Detection (VAD)](https://docs.runanywhere.ai/flutter/introduction#voice-activity-detection-vad)
    
*   [Voice Agent Pipeline](https://docs.runanywhere.ai/flutter/introduction#voice-agent-pipeline)
    
*   [System Requirements](https://docs.runanywhere.ai/flutter/introduction#system-requirements)
    
*   [Package Composition](https://docs.runanywhere.ai/flutter/introduction#package-composition)
    
*   [Architecture](https://docs.runanywhere.ai/flutter/introduction#architecture)
    
*   [Starter Example](https://docs.runanywhere.ai/flutter/introduction#starter-example)
    
*   [Next Steps](https://docs.runanywhere.ai/flutter/introduction#next-steps)
    

[​](https://docs.runanywhere.ai/flutter/introduction#overview)

Overview
--------------------------------------------------------------------------

The **RunAnywhere Flutter SDK** is a production-grade, on-device AI SDK for Flutter applications. It enables developers to run AI models directly on iOS and Android devices without requiring network connectivity for inference, ensuring minimal latency and maximum privacy for your users. The SDK provides a unified interface to multiple AI capabilities:

[LLM\
---\
\
Text generation with streaming support via Dart Streams](https://docs.runanywhere.ai/flutter/llm/chat)
[STT\
---\
\
Speech-to-text transcription with Whisper models](https://docs.runanywhere.ai/flutter/stt/transcribe)
[TTS\
---\
\
Neural voice synthesis with Piper TTS](https://docs.runanywhere.ai/flutter/tts/synthesize)
[VAD\
---\
\
Real-time voice activity detection with Silero VAD](https://docs.runanywhere.ai/flutter/vad)

[​](https://docs.runanywhere.ai/flutter/introduction#key-capabilities)

Key Capabilities
------------------------------------------------------------------------------------------

*   **Multi-backend architecture** – Choose from LlamaCPP (GGUF models) or ONNX Runtime
*   **Cross-platform** – Single codebase for iOS and Android
*   **Dart-native** – Built with async/await and Streams for reactive programming
*   **Production-ready** – Built-in analytics, logging, and model lifecycle management

[​](https://docs.runanywhere.ai/flutter/introduction#core-philosophy)

Core Philosophy
----------------------------------------------------------------------------------------

On-Device First

All AI inference runs locally, ensuring low latency and data privacy. Once models are downloaded, no network connection is required for inference.

Modular Backends

Backend engines are separate packages—include only what you need. This keeps your app bundle size minimal.

Privacy by Design

Audio and text data never leaves the device unless explicitly configured. Only anonymous analytics are collected by default.

Event-Driven

Subscribe to SDK events for reactive UI updates and observability.

[​](https://docs.runanywhere.ai/flutter/introduction#features)

Features
--------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/flutter/introduction#language-models-llm)

Language Models (LLM)

*   On-device text generation with streaming support
*   Dart Stream-based token streaming
*   System prompts and customizable generation parameters
*   Support for thinking/reasoning models
*   LlamaCPP backend for GGUF models

### 

[​](https://docs.runanywhere.ai/flutter/introduction#speech-to-text-stt)

Speech-to-Text (STT)

*   Real-time streaming transcription
*   Batch audio transcription
*   Multi-language support
*   Whisper-based models via ONNX Runtime

### 

[​](https://docs.runanywhere.ai/flutter/introduction#text-to-speech-tts)

Text-to-Speech (TTS)

*   Neural voice synthesis with Piper TTS
*   System voices via platform TTS
*   Streaming audio generation for long text
*   Customizable voice, pitch, rate, and volume

### 

[​](https://docs.runanywhere.ai/flutter/introduction#voice-activity-detection-vad)

Voice Activity Detection (VAD)

*   Energy-based speech detection with Silero VAD
*   Configurable sensitivity thresholds
*   Real-time audio stream processing

### 

[​](https://docs.runanywhere.ai/flutter/introduction#voice-agent-pipeline)

Voice Agent Pipeline

*   Full VAD → STT → LLM → TTS orchestration
*   Complete voice conversation flow
*   Push-to-talk and hands-free modes

[​](https://docs.runanywhere.ai/flutter/introduction#system-requirements)

System Requirements
------------------------------------------------------------------------------------------------

| Platform | Minimum Version |
| --- | --- |
| Flutter | 3.10.0+ |
| Dart | 3.0.0+ |
| iOS | 14.0+ |
| Android | API 24 (7.0+) |

ARM64 devices are recommended for best performance. Metal GPU acceleration on iOS and NEON SIMD on Android provide significant speedups over CPU-only inference.

[​](https://docs.runanywhere.ai/flutter/introduction#package-composition)

Package Composition
------------------------------------------------------------------------------------------------

| Package | Size | Purpose |
| --- | --- | --- |
| `runanywhere` | ~5MB | Core SDK (required) |
| `runanywhere_llamacpp` | ~15-25MB | LLM text generation with GGUF models |
| `runanywhere_onnx` | ~50-70MB | STT/TTS/VAD via ONNX Runtime |

[​](https://docs.runanywhere.ai/flutter/introduction#architecture)

Architecture
----------------------------------------------------------------------------------

[​](https://docs.runanywhere.ai/flutter/introduction#starter-example)

Starter Example
----------------------------------------------------------------------------------------

[Flutter Starter Example\
-----------------------\
\
Complete working example with LLM chat, STT, TTS, and Voice Agent demos](https://github.com/RunanywhereAI/flutter-starter-example)

[​](https://docs.runanywhere.ai/flutter/introduction#next-steps)

Next Steps
------------------------------------------------------------------------------

[Installation\
------------\
\
Add the SDK packages via pub.dev](https://docs.runanywhere.ai/flutter/installation)
[Quick Start\
-----------\
\
Build your first AI feature in minutes](https://docs.runanywhere.ai/flutter/quick-start)

[Best Practices](https://docs.runanywhere.ai/react-native/best-practices)
[Installation](https://docs.runanywhere.ai/flutter/installation)

⌘I

---

# chat() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/llm/chat#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

chat()

On this page

*   [When to Use](https://docs.runanywhere.ai/flutter/llm/chat#when-to-use)
    
*   [Requirements](https://docs.runanywhere.ai/flutter/llm/chat#requirements)
    
*   [Example with Error Handling](https://docs.runanywhere.ai/flutter/llm/chat#example-with-error-handling)
    
*   [See Also](https://docs.runanywhere.ai/flutter/llm/chat#see-also)
    

The simplest way to generate text—returns just the response string.

Copy

Ask AI

    // One-liner for quick responses
    final response = await RunAnywhere.chat('What is the capital of France?');
    print(response);  // "The capital of France is Paris."
    

[​](https://docs.runanywhere.ai/flutter/llm/chat#when-to-use)

When to Use
----------------------------------------------------------------------------

Use `chat()` when you:

*   Need a quick response without metrics
*   Don’t need streaming
*   Want the simplest possible API

[​](https://docs.runanywhere.ai/flutter/llm/chat#requirements)

Requirements
------------------------------------------------------------------------------

Before calling `chat()`, ensure:

1.  SDK is initialized: `await RunAnywhere.initialize()`
2.  Backend is registered: `await LlamaCpp.register()`
3.  Model is loaded: `await RunAnywhere.loadModel('model-id')`

[​](https://docs.runanywhere.ai/flutter/llm/chat#example-with-error-handling)

Example with Error Handling
------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    Future<String> askQuestion(String question) async {
      try {
        if (!RunAnywhere.isModelLoaded) {
          throw SDKError.componentNotReady('No model loaded');
        }
        return await RunAnywhere.chat(question);
      } on SDKError catch (e) {
        return 'Error: ${e.message}';
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/llm/chat#see-also)

See Also
----------------------------------------------------------------------

[generate()\
----------\
\
Get response with metrics](https://docs.runanywhere.ai/flutter/llm/generate)
[generateStream()\
----------------\
\
Stream tokens in real-time](https://docs.runanywhere.ai/flutter/llm/stream)

[Quick Start](https://docs.runanywhere.ai/flutter/quick-start)
[generate()](https://docs.runanywhere.ai/flutter/llm/generate)

⌘I

---

# Streaming TTS - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/tts/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

Streaming TTS

On this page

*   [Overview](https://docs.runanywhere.ai/flutter/tts/stream#overview)
    
*   [Basic Concept](https://docs.runanywhere.ai/flutter/tts/stream#basic-concept)
    
*   [Chunked Synthesis](https://docs.runanywhere.ai/flutter/tts/stream#chunked-synthesis)
    
*   [With Voice Agent Pipeline](https://docs.runanywhere.ai/flutter/tts/stream#with-voice-agent-pipeline)
    
*   [Latency Optimization Tips](https://docs.runanywhere.ai/flutter/tts/stream#latency-optimization-tips)
    
*   [See Also](https://docs.runanywhere.ai/flutter/tts/stream#see-also)
    

Stream TTS audio as it’s generated for faster time-to-first-audio, especially with longer text.

[​](https://docs.runanywhere.ai/flutter/tts/stream#overview)

Overview
------------------------------------------------------------------------

Streaming TTS starts playing audio before the entire synthesis is complete. This is particularly useful for:

*   Long text passages
*   Voice assistants responding in real-time
*   Reducing perceived latency

[​](https://docs.runanywhere.ai/flutter/tts/stream#basic-concept)

Basic Concept
----------------------------------------------------------------------------------

Copy

Ask AI

    // The Voice Agent pipeline handles streaming TTS automatically
    final session = await RunAnywhere.startVoiceSession(
      config: VoiceSessionConfig(
        autoPlayTTS: true,  // Automatically plays synthesized audio
      ),
    );
    
    session.events.listen((event) {
      if (event is VoiceSessionSpeaking) {
        print('Playing audio response...');
      }
    });
    

[​](https://docs.runanywhere.ai/flutter/tts/stream#chunked-synthesis)

Chunked Synthesis
------------------------------------------------------------------------------------------

For manual control, synthesize text in chunks:

Copy

Ask AI

    Future<void> speakInChunks(String longText) async {
      // Split into sentences
      final sentences = longText.split(RegExp(r'(?<=[.!?])\s+'));
    
      for (final sentence in sentences) {
        final result = await RunAnywhere.synthesize(sentence);
        await playAudio(result);
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/tts/stream#with-voice-agent-pipeline)

With Voice Agent Pipeline
----------------------------------------------------------------------------------------------------------

The Voice Agent provides the best streaming TTS experience:

Copy

Ask AI

    // Initialize all components
    await RunAnywhere.loadSTTModel('sherpa-onnx-whisper-tiny.en');
    await RunAnywhere.loadModel('smollm2-360m-q8_0');
    await RunAnywhere.loadTTSVoice('vits-piper-en_US-lessac-medium');
    
    // Start session with auto-play
    final session = await RunAnywhere.startVoiceSession(
      config: VoiceSessionConfig(
        autoPlayTTS: true,
        continuousMode: true,
      ),
    );
    
    // The pipeline automatically:
    // 1. Detects speech (VAD)
    // 2. Transcribes audio (STT)
    // 3. Generates response (LLM)
    // 4. Synthesizes and plays audio (TTS)
    session.events.listen((event) {
      switch (event) {
        case VoiceSessionTranscribed(:final text):
          print('User: $text');
        case VoiceSessionResponded(:final text):
          print('AI: $text');
        case VoiceSessionSpeaking():
          print('Playing response...');
        case VoiceSessionTurnCompleted():
          print('Ready for next turn');
      }
    });
    

[​](https://docs.runanywhere.ai/flutter/tts/stream#latency-optimization-tips)

Latency Optimization Tips
----------------------------------------------------------------------------------------------------------

Preload Voice Model

Load the TTS voice during app startup or idle time, not when the user first needs it.

Copy

Ask AI

    // In app initialization
    await RunAnywhere.loadTTSVoice('vits-piper-en_US-lessac-medium');

Use Faster Voices

Smaller voice models synthesize faster. Choose based on your quality/speed tradeoff.

Sentence-Level Streaming

For very long responses, synthesize and play sentence by sentence rather than waiting for the complete response.

Buffer Management

Start playback as soon as you have enough audio buffered (typically 100-200ms).

[​](https://docs.runanywhere.ai/flutter/tts/stream#see-also)

See Also
------------------------------------------------------------------------

[synthesize()\
------------\
\
Basic synthesis](https://docs.runanywhere.ai/flutter/tts/synthesize)
[Voice Agent\
-----------\
\
Complete voice pipeline](https://docs.runanywhere.ai/flutter/voice-agent)

[synthesize()](https://docs.runanywhere.ai/flutter/tts/synthesize)
[Voices](https://docs.runanywhere.ai/flutter/tts/voices)

⌘I

---

# generate() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/llm/generate#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

generate()

On this page

*   [LLMGenerationOptions](https://docs.runanywhere.ai/flutter/llm/generate#llmgenerationoptions)
    
*   [LLMGenerationResult](https://docs.runanywhere.ai/flutter/llm/generate#llmgenerationresult)
    
*   [Thinking Models](https://docs.runanywhere.ai/flutter/llm/generate#thinking-models)
    
*   [See Also](https://docs.runanywhere.ai/flutter/llm/generate#see-also)
    

Generate text with detailed metrics including latency, token count, and generation speed.

Copy

Ask AI

    final result = await RunAnywhere.generate(
      'Explain quantum computing in simple terms',
      options: LLMGenerationOptions(
        maxTokens: 200,
        temperature: 0.7,
      ),
    );
    
    print('Response: ${result.text}');
    print('Tokens: ${result.tokensUsed}');
    print('Speed: ${result.tokensPerSecond.toStringAsFixed(1)} tok/s');
    print('Latency: ${result.latencyMs.toStringAsFixed(0)}ms');
    

[​](https://docs.runanywhere.ai/flutter/llm/generate#llmgenerationoptions)

LLMGenerationOptions
--------------------------------------------------------------------------------------------------

| Parameter | Type | Default | Description |
| --- | --- | --- | --- |
| `maxTokens` | `int` | 100 | Maximum tokens to generate |
| `temperature` | `double` | 0.8 | Randomness (0.0–2.0) |
| `topP` | `double` | 1.0 | Nucleus sampling parameter |
| `stopSequences` | `List<String>` | `[]` | Stop generation at these |
| `systemPrompt` | `String?` | `null` | System prompt for context |

Copy

Ask AI

    const options = LLMGenerationOptions(
      maxTokens: 256,
      temperature: 0.7,
      topP: 0.95,
      stopSequences: ['END', '###'],
      systemPrompt: 'You are a helpful coding assistant.',
    );
    

[​](https://docs.runanywhere.ai/flutter/llm/generate#llmgenerationresult)

LLMGenerationResult
------------------------------------------------------------------------------------------------

| Property | Type | Description |
| --- | --- | --- |
| `text` | `String` | Generated text |
| `thinkingContent` | `String?` | Thinking content (if supported) |
| `inputTokens` | `int` | Number of input tokens |
| `tokensUsed` | `int` | Number of output tokens |
| `modelUsed` | `String` | Model ID used |
| `latencyMs` | `double` | Total latency in milliseconds |
| `tokensPerSecond` | `double` | Generation speed |
| `timeToFirstTokenMs` | `double?` | Time to first token |

[​](https://docs.runanywhere.ai/flutter/llm/generate#thinking-models)

Thinking Models
----------------------------------------------------------------------------------------

Some models support “thinking” tokens for chain-of-thought reasoning:

Copy

Ask AI

    LlamaCpp.addModel(
      id: 'qwen-cot',
      name: 'Qwen CoT',
      url: '...',
      supportsThinking: true,  // Enable thinking token parsing
    );
    
    final result = await RunAnywhere.generate('Solve: 2x + 5 = 15');
    
    if (result.thinkingContent != null) {
      print('Reasoning: ${result.thinkingContent}');
    }
    print('Answer: ${result.text}');
    

[​](https://docs.runanywhere.ai/flutter/llm/generate#see-also)

See Also
--------------------------------------------------------------------------

[chat()\
------\
\
Simple one-liner](https://docs.runanywhere.ai/flutter/llm/chat)
[generateStream()\
----------------\
\
Stream tokens in real-time](https://docs.runanywhere.ai/flutter/llm/stream)

[chat()](https://docs.runanywhere.ai/flutter/llm/chat)
[generateStream()](https://docs.runanywhere.ai/flutter/llm/stream)

⌘I

---

# Streaming STT - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/stt/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

Streaming STT

On this page

*   [Overview](https://docs.runanywhere.ai/flutter/stt/stream#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/flutter/stt/stream#basic-usage)
    
*   [With Voice Session](https://docs.runanywhere.ai/flutter/stt/stream#with-voice-session)
    
*   [Real-Time Transcription Widget](https://docs.runanywhere.ai/flutter/stt/stream#real-time-transcription-widget)
    
*   [Tips for Streaming STT](https://docs.runanywhere.ai/flutter/stt/stream#tips-for-streaming-stt)
    
*   [See Also](https://docs.runanywhere.ai/flutter/stt/stream#see-also)
    

Stream audio for real-time transcription as the user speaks. Ideal for voice assistants and live captioning.

[​](https://docs.runanywhere.ai/flutter/stt/stream#overview)

Overview
------------------------------------------------------------------------

Streaming STT processes audio in chunks, providing partial transcriptions that update as more audio arrives. This creates a responsive experience where users see their words appear in real-time.

[​](https://docs.runanywhere.ai/flutter/stt/stream#basic-usage)

Basic Usage
------------------------------------------------------------------------------

Copy

Ask AI

    // Ensure STT model is loaded
    await RunAnywhere.loadSTTModel('sherpa-onnx-whisper-tiny.en');
    
    // Create audio stream from microphone
    final audioStream = microphoneManager.audioStream;
    
    // Process audio chunks
    await for (final audioChunk in audioStream) {
      // Feed audio to STT (returns partial transcription)
      final partial = await RunAnywhere.transcribeChunk(audioChunk);
      print('Partial: $partial');
    }
    

[​](https://docs.runanywhere.ai/flutter/stt/stream#with-voice-session)

With Voice Session
--------------------------------------------------------------------------------------------

The easiest way to use streaming STT is through the Voice Agent, which handles VAD and audio capture automatically:

Copy

Ask AI

    final session = await RunAnywhere.startVoiceSession();
    
    session.events.listen((event) {
      if (event is VoiceSessionTranscribed) {
        print('Transcription: ${event.text}');
      }
    });
    

See [Voice Agent](https://docs.runanywhere.ai/flutter/voice-agent)
 for the complete voice pipeline.

[​](https://docs.runanywhere.ai/flutter/stt/stream#real-time-transcription-widget)

Real-Time Transcription Widget
--------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class LiveTranscriptionWidget extends StatefulWidget {
      @override
      _LiveTranscriptionWidgetState createState() => _LiveTranscriptionWidgetState();
    }
    
    class _LiveTranscriptionWidgetState extends State<LiveTranscriptionWidget> {
      final _recorder = AudioRecorder();
      String _partialText = '';
      String _finalText = '';
      bool _isListening = false;
      StreamSubscription? _audioSubscription;
    
      Future<void> _startListening() async {
        if (!await _recorder.hasPermission()) return;
    
        setState(() {
          _isListening = true;
          _partialText = '';
        });
    
        // Start recording with streaming
        final stream = await _recorder.startStream(
          const RecordConfig(
            encoder: AudioEncoder.pcm16bits,
            sampleRate: 16000,
            numChannels: 1,
          ),
        );
    
        // Process audio chunks
        _audioSubscription = stream.listen((chunk) async {
          // This would use your streaming transcription implementation
          // The exact API depends on how you've set up streaming
        });
      }
    
      Future<void> _stopListening() async {
        await _audioSubscription?.cancel();
        await _recorder.stop();
    
        setState(() {
          _isListening = false;
          _finalText = _partialText;
          _partialText = '';
        });
      }
    
      @override
      Widget build(BuildContext context) {
        return Column(
          children: [\
            // Show partial transcription with typing indicator\
            Container(\
              padding: EdgeInsets.all(16),\
              child: Text(\
                _isListening ? '$_partialText|' : _finalText,\
                style: TextStyle(\
                  fontSize: 18,\
                  color: _isListening ? Colors.grey : Colors.black,\
                ),\
              ),\
            ),\
    \
            // Recording button\
            IconButton(\
              icon: Icon(_isListening ? Icons.stop : Icons.mic),\
              iconSize: 48,\
              color: _isListening ? Colors.red : Colors.blue,\
              onPressed: _isListening ? _stopListening : _startListening,\
            ),\
          ],
        );
      }
    
      @override
      void dispose() {
        _audioSubscription?.cancel();
        _recorder.dispose();
        super.dispose();
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/stt/stream#tips-for-streaming-stt)

Tips for Streaming STT
----------------------------------------------------------------------------------------------------

Buffer Management

Accumulate audio in buffers of 100-500ms for optimal accuracy vs latency tradeoff.

Silence Detection

Use VAD (Voice Activity Detection) to detect end of speech and finalize transcriptions.

Error Recovery

Handle network interruptions and audio glitches gracefully. Consider retrying failed chunks.

UI Feedback

Show a visual indicator (waveform, pulsing dot) to confirm audio is being captured.

[​](https://docs.runanywhere.ai/flutter/stt/stream#see-also)

See Also
------------------------------------------------------------------------

[transcribe()\
------------\
\
Batch transcription](https://docs.runanywhere.ai/flutter/stt/transcribe)
[Voice Agent\
-----------\
\
Complete voice pipeline](https://docs.runanywhere.ai/flutter/voice-agent)

[transcribe()](https://docs.runanywhere.ai/flutter/stt/transcribe)
[STT Options](https://docs.runanywhere.ai/flutter/stt/options)

⌘I

---

# Configuration - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/configuration#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Kotlin SDK

Configuration

On this page

*   [SDK Initialization](https://docs.runanywhere.ai/kotlin/configuration#sdk-initialization)
    
*   [SDKEnvironment](https://docs.runanywhere.ai/kotlin/configuration#sdkenvironment)
    
*   [LLM Generation Options](https://docs.runanywhere.ai/kotlin/configuration#llm-generation-options)
    
*   [Temperature Guide](https://docs.runanywhere.ai/kotlin/configuration#temperature-guide)
    
*   [STT Options](https://docs.runanywhere.ai/kotlin/configuration#stt-options)
    
*   [TTS Options](https://docs.runanywhere.ai/kotlin/configuration#tts-options)
    
*   [VAD Configuration](https://docs.runanywhere.ai/kotlin/configuration#vad-configuration)
    
*   [Memory Considerations](https://docs.runanywhere.ai/kotlin/configuration#memory-considerations)
    
*   [Model Registry](https://docs.runanywhere.ai/kotlin/configuration#model-registry)
    
*   [InferenceFramework](https://docs.runanywhere.ai/kotlin/configuration#inferenceframework)
    
*   [ModelCategory](https://docs.runanywhere.ai/kotlin/configuration#modelcategory)
    
*   [SDK Lifecycle](https://docs.runanywhere.ai/kotlin/configuration#sdk-lifecycle)
    

[​](https://docs.runanywhere.ai/kotlin/configuration#sdk-initialization)

SDK Initialization
----------------------------------------------------------------------------------------------

Copy

Ask AI

    RunAnywhere.initialize(
        apiKey = "your-api-key",           // Optional for development
        baseURL = "https://api.runanywhere.ai",  // Custom API endpoint
        environment = SDKEnvironment.DEVELOPMENT
    )
    

[​](https://docs.runanywhere.ai/kotlin/configuration#sdkenvironment)

SDKEnvironment
--------------------------------------------------------------------------------------

| Environment | Log Level | Description |
| --- | --- | --- |
| `DEVELOPMENT` | Debug | Full logging, local testing |
| `STAGING` | Info | Staging backend, moderate logging |
| `PRODUCTION` | Warning | Production backend, minimal logging |

[​](https://docs.runanywhere.ai/kotlin/configuration#llm-generation-options)

LLM Generation Options
------------------------------------------------------------------------------------------------------

Copy

Ask AI

    data class LLMGenerationOptions(
        val maxTokens: Int = 100,           // Maximum tokens to generate
        val temperature: Float = 0.8f,      // Creativity (0.0-2.0)
        val topP: Float = 1.0f,             // Nucleus sampling (0.0-1.0)
        val stopSequences: List<String> = emptyList(),
        val streamingEnabled: Boolean = false,
        val preferredFramework: InferenceFramework? = null,
        val systemPrompt: String? = null
    )
    

### 

[​](https://docs.runanywhere.ai/kotlin/configuration#temperature-guide)

Temperature Guide

| Value | Use Case | Description |
| --- | --- | --- |
| 0.0-0.3 | Code, facts | Deterministic, consistent |
| 0.4-0.7 | General Q&A | Balanced |
| 0.8-1.0 | Creative writing | More variety |
| 1.1-2.0 | Brainstorming | High creativity |

[​](https://docs.runanywhere.ai/kotlin/configuration#stt-options)

STT Options
--------------------------------------------------------------------------------

Copy

Ask AI

    data class STTOptions(
        val language: String = "en",
        val detectLanguage: Boolean = false,
        val enablePunctuation: Boolean = true,
        val enableDiarization: Boolean = false,
        val maxSpeakers: Int? = null,
        val enableTimestamps: Boolean = true,
        val vocabularyFilter: List<String> = emptyList(),
        val audioFormat: AudioFormat = AudioFormat.PCM,
        val sampleRate: Int = 16000
    )
    

[​](https://docs.runanywhere.ai/kotlin/configuration#tts-options)

TTS Options
--------------------------------------------------------------------------------

Copy

Ask AI

    data class TTSOptions(
        val voice: String? = null,
        val language: String = "en-US",
        val rate: Float = 1.0f,          // 0.0-2.0
        val pitch: Float = 1.0f,         // 0.0-2.0
        val volume: Float = 1.0f,        // 0.0-1.0
        val audioFormat: AudioFormat = AudioFormat.PCM,
        val sampleRate: Int = 22050,
        val useSSML: Boolean = false
    )
    

[​](https://docs.runanywhere.ai/kotlin/configuration#vad-configuration)

VAD Configuration
--------------------------------------------------------------------------------------------

Copy

Ask AI

    data class VADConfiguration(
        val threshold: Float = 0.5f,
        val minSpeechDurationMs: Int = 250,
        val minSilenceDurationMs: Int = 300,
        val sampleRate: Int = 16000,
        val frameSizeMs: Int = 30
    )
    

[​](https://docs.runanywhere.ai/kotlin/configuration#memory-considerations)

Memory Considerations
----------------------------------------------------------------------------------------------------

Model memory requirements vary by size and quantization:

| Model Type | Memory | Device Requirements |
| --- | --- | --- |
| LLM 0.5B Q8 | ~500MB | Any Android 7.0+ |
| LLM 0.5B Q4 | ~300MB | Any Android 7.0+ |
| LLM 1B Q4 | ~600MB | 4GB+ RAM |
| LLM 3B Q4 | ~2GB | 6GB+ RAM |
| Whisper Tiny | ~75MB | Any device |
| Whisper Base | ~150MB | Any device |
| TTS Voice | ~50MB | Any device |

Always check available memory before loading large models. Loading models that exceed available RAM will crash the app.

[​](https://docs.runanywhere.ai/kotlin/configuration#model-registry)

Model Registry
--------------------------------------------------------------------------------------

Register custom models:

Copy

Ask AI

    val modelInfo = RunAnywhere.registerModel(
        id = "my-custom-model",           // Optional, generated from URL if null
        name = "My Custom Model",
        url = "https://...",
        framework = InferenceFramework.LLAMA_CPP,
        modality = ModelCategory.LANGUAGE,
        memoryRequirement = 500_000_000L,  // 500MB
        supportsThinking = false
    )
    

[​](https://docs.runanywhere.ai/kotlin/configuration#inferenceframework)

InferenceFramework
----------------------------------------------------------------------------------------------

Copy

Ask AI

    enum class InferenceFramework {
        ONNX,              // ONNX Runtime (STT/TTS/VAD)
        LLAMA_CPP,         // llama.cpp (LLM)
        FOUNDATION_MODELS, // Platform foundation models
        SYSTEM_TTS,        // System text-to-speech
        BUILT_IN,          // Simple built-in services
        NONE,              // No model needed
        UNKNOWN
    }
    

[​](https://docs.runanywhere.ai/kotlin/configuration#modelcategory)

ModelCategory
------------------------------------------------------------------------------------

Copy

Ask AI

    enum class ModelCategory {
        LANGUAGE,              // LLMs (text-to-text)
        SPEECH_RECOGNITION,    // STT (voice-to-text)
        SPEECH_SYNTHESIS,      // TTS (text-to-voice)
        VISION,                // Image understanding
        IMAGE_GENERATION,      // Text-to-image
        MULTIMODAL,            // Multiple modalities
        AUDIO                  // Audio processing
    }
    

[​](https://docs.runanywhere.ai/kotlin/configuration#sdk-lifecycle)

SDK Lifecycle
------------------------------------------------------------------------------------

Copy

Ask AI

    // Check initialization state
    val isInit = RunAnywhere.isInitialized
    val servicesReady = RunAnywhere.areServicesReady
    
    // Get SDK version
    val version = RunAnywhere.version
    
    // Get current environment
    val env = RunAnywhere.environment
    
    // Reset SDK (clear all state)
    RunAnywhere.reset()
    
    // Cleanup resources without full reset
    RunAnywhere.cleanup()
    

[Voice Agent](https://docs.runanywhere.ai/kotlin/voice-agent)
[Error Handling](https://docs.runanywhere.ai/kotlin/error-handling)

⌘I

---

# generateStream() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/llm/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

generateStream()

On this page

*   [LLMStreamingResult](https://docs.runanywhere.ai/flutter/llm/stream#llmstreamingresult)
    
*   [Flutter Widget Example](https://docs.runanywhere.ai/flutter/llm/stream#flutter-widget-example)
    
*   [Cancellation](https://docs.runanywhere.ai/flutter/llm/stream#cancellation)
    
*   [Best Practices](https://docs.runanywhere.ai/flutter/llm/stream#best-practices)
    
*   [See Also](https://docs.runanywhere.ai/flutter/llm/stream#see-also)
    

Stream tokens in real-time for a responsive user experience. Ideal for chat interfaces where users expect to see text appear progressively.

Copy

Ask AI

    final streamResult = await RunAnywhere.generateStream(
      'Tell me a story about a robot',
      options: LLMGenerationOptions(maxTokens: 500),
    );
    
    // Display tokens as they arrive
    await for (final token in streamResult.stream) {
      stdout.write(token);  // Real-time output
      // Or update UI: setState(() => _response += token);
    }
    
    // Get final metrics after streaming completes
    final metrics = await streamResult.result;
    print('\n\nGenerated ${metrics.tokensUsed} tokens');
    print('Speed: ${metrics.tokensPerSecond.toStringAsFixed(1)} tok/s');
    

[​](https://docs.runanywhere.ai/flutter/llm/stream#llmstreamingresult)

LLMStreamingResult
--------------------------------------------------------------------------------------------

| Property | Type | Description |
| --- | --- | --- |
| `stream` | `Stream<String>` | Stream of tokens |
| `result` | `Future<LLMGenerationResult>` | Final result with metrics |
| `cancel` | `void Function()` | Cancel generation |

[​](https://docs.runanywhere.ai/flutter/llm/stream#flutter-widget-example)

Flutter Widget Example
----------------------------------------------------------------------------------------------------

Copy

Ask AI

    class StreamingChatWidget extends StatefulWidget {
      @override
      _StreamingChatWidgetState createState() => _StreamingChatWidgetState();
    }
    
    class _StreamingChatWidgetState extends State<StreamingChatWidget> {
      String _response = '';
      bool _isStreaming = false;
      LLMStreamingResult? _currentStream;
    
      Future<void> _startStreaming(String prompt) async {
        setState(() {
          _response = '';
          _isStreaming = true;
        });
    
        try {
          _currentStream = await RunAnywhere.generateStream(
            prompt,
            options: LLMGenerationOptions(maxTokens: 300),
          );
    
          await for (final token in _currentStream!.stream) {
            setState(() => _response += token);
          }
    
          final metrics = await _currentStream!.result;
          print('Generated at ${metrics.tokensPerSecond.toStringAsFixed(1)} tok/s');
        } finally {
          setState(() => _isStreaming = false);
        }
      }
    
      void _cancelStreaming() {
        _currentStream?.cancel();
      }
    
      @override
      Widget build(BuildContext context) {
        return Column(
          children: [\
            Expanded(\
              child: SingleChildScrollView(\
                child: Text(_response),\
              ),\
            ),\
            Row(\
              children: [\
                ElevatedButton(\
                  onPressed: _isStreaming ? null : () => _startStreaming('Tell me a joke'),\
                  child: Text('Generate'),\
                ),\
                if (_isStreaming)\
                  TextButton(\
                    onPressed: _cancelStreaming,\
                    child: Text('Cancel'),\
                  ),\
              ],\
            ),\
          ],
        );
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/llm/stream#cancellation)

Cancellation
--------------------------------------------------------------------------------

Cancel ongoing generation at any time:

Copy

Ask AI

    final streamResult = await RunAnywhere.generateStream('Long story...');
    
    // Start a timer to cancel after 5 seconds
    Future.delayed(Duration(seconds: 5), () {
      streamResult.cancel();
      print('Generation cancelled');
    });
    
    // Or cancel via the static method
    await RunAnywhere.cancelGeneration();
    

[​](https://docs.runanywhere.ai/flutter/llm/stream#best-practices)

Best Practices
------------------------------------------------------------------------------------

**Use streaming for chat interfaces** — Users perceive the app as more responsive when they see tokens appear progressively, even if total generation time is the same.

1.  **Update UI incrementally** — Append tokens to your state as they arrive
2.  **Show a cancel button** — Let users stop long generations
3.  **Handle cancellation gracefully** — The stream will complete when cancelled
4.  **Get final metrics** — Always await `result` for accurate performance data

[​](https://docs.runanywhere.ai/flutter/llm/stream#see-also)

See Also
------------------------------------------------------------------------

[chat()\
------\
\
Simple one-liner](https://docs.runanywhere.ai/flutter/llm/chat)
[generate()\
----------\
\
Non-streaming with metrics](https://docs.runanywhere.ai/flutter/llm/generate)

[generate()](https://docs.runanywhere.ai/flutter/llm/generate)
[System Prompts](https://docs.runanywhere.ai/flutter/llm/system-prompts)

⌘I

---

# Voices - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/tts/voices#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

Voices

On this page

*   [Piper Neural Voices](https://docs.runanywhere.ai/flutter/tts/voices#piper-neural-voices)
    
*   [English (US)](https://docs.runanywhere.ai/flutter/tts/voices#english-us)
    
*   [English (UK)](https://docs.runanywhere.ai/flutter/tts/voices#english-uk)
    
*   [Other Languages](https://docs.runanywhere.ai/flutter/tts/voices#other-languages)
    
*   [Registering Voices](https://docs.runanywhere.ai/flutter/tts/voices#registering-voices)
    
*   [Switching Voices](https://docs.runanywhere.ai/flutter/tts/voices#switching-voices)
    
*   [Voice Selection UI](https://docs.runanywhere.ai/flutter/tts/voices#voice-selection-ui)
    
*   [System Voice Fallback](https://docs.runanywhere.ai/flutter/tts/voices#system-voice-fallback)
    
*   [Choosing Between Neural and System Voices](https://docs.runanywhere.ai/flutter/tts/voices#choosing-between-neural-and-system-voices)
    
*   [See Also](https://docs.runanywhere.ai/flutter/tts/voices#see-also)
    

RunAnywhere supports neural TTS voices via Piper and system voice fallback via `flutter_tts`.

[​](https://docs.runanywhere.ai/flutter/tts/voices#piper-neural-voices)

Piper Neural Voices
----------------------------------------------------------------------------------------------

High-quality neural voices that run entirely on-device.

### 

[​](https://docs.runanywhere.ai/flutter/tts/voices#english-us)

English (US)

| Voice ID | Name | Gender | Style | Size |
| --- | --- | --- | --- | --- |
| `vits-piper-en_US-lessac-medium` | Lessac | Male | Neutral | ~50MB |
| `vits-piper-en_US-amy-medium` | Amy | Female | Neutral | ~50MB |
| `vits-piper-en_US-ryan-medium` | Ryan | Male | Conversational | ~50MB |
| `vits-piper-en_US-kathleen-low` | Kathleen | Female | Soft | ~20MB |

### 

[​](https://docs.runanywhere.ai/flutter/tts/voices#english-uk)

English (UK)

| Voice ID | Name | Gender | Style | Size |
| --- | --- | --- | --- | --- |
| `vits-piper-en_GB-alan-medium` | Alan | Male | British | ~50MB |
| `vits-piper-en_GB-alba-medium` | Alba | Female | Scottish | ~50MB |

### 

[​](https://docs.runanywhere.ai/flutter/tts/voices#other-languages)

Other Languages

| Voice ID | Language | Gender | Size |
| --- | --- | --- | --- |
| `vits-piper-de_DE-thorsten-medium` | German | Male | ~50MB |
| `vits-piper-fr_FR-upmc-medium` | French | Male | ~50MB |
| `vits-piper-es_ES-sharvard-medium` | Spanish | Male | ~50MB |
| `vits-piper-it_IT-riccardo-x_low` | Italian | Male | ~20MB |
| `vits-piper-zh_CN-huayan-medium` | Chinese | Female | ~50MB |

[​](https://docs.runanywhere.ai/flutter/tts/voices#registering-voices)

Registering Voices
--------------------------------------------------------------------------------------------

Copy

Ask AI

    // US English - Lessac (recommended default)
    Onnx.addModel(
      id: 'vits-piper-en_US-lessac-medium',
      name: 'Piper US English (Lessac)',
      url: 'https://github.com/RunanywhereAI/sherpa-onnx/releases/download/runanywhere-models-v1/vits-piper-en_US-lessac-medium.tar.bz2',
      modality: ModelCategory.textToSpeech,
    );
    
    // US English - Amy (female)
    Onnx.addModel(
      id: 'vits-piper-en_US-amy-medium',
      name: 'Piper US English (Amy)',
      url: 'https://github.com/RunanywhereAI/sherpa-onnx/releases/download/runanywhere-models-v1/vits-piper-en_US-amy-medium.tar.bz2',
      modality: ModelCategory.textToSpeech,
    );
    

[​](https://docs.runanywhere.ai/flutter/tts/voices#switching-voices)

Switching Voices
----------------------------------------------------------------------------------------

Copy

Ask AI

    // Load default voice
    await RunAnywhere.loadTTSVoice('vits-piper-en_US-lessac-medium');
    
    // Switch to different voice
    Future<void> switchVoice(String voiceId) async {
      await RunAnywhere.unloadTTSVoice();
      await RunAnywhere.loadTTSVoice(voiceId);
    }
    

[​](https://docs.runanywhere.ai/flutter/tts/voices#voice-selection-ui)

Voice Selection UI
--------------------------------------------------------------------------------------------

Copy

Ask AI

    class VoiceSelectorWidget extends StatefulWidget {
      @override
      _VoiceSelectorWidgetState createState() => _VoiceSelectorWidgetState();
    }
    
    class _VoiceSelectorWidgetState extends State<VoiceSelectorWidget> {
      final voices = [\
        {'id': 'vits-piper-en_US-lessac-medium', 'name': 'Lessac (Male)'},\
        {'id': 'vits-piper-en_US-amy-medium', 'name': 'Amy (Female)'},\
        {'id': 'vits-piper-en_US-ryan-medium', 'name': 'Ryan (Male)'},\
      ];
    
      String? _selectedVoice;
      bool _isLoading = false;
    
      Future<void> _selectVoice(String voiceId) async {
        setState(() => _isLoading = true);
    
        try {
          // Download if needed
          await for (final p in RunAnywhere.downloadModel(voiceId)) {
            if (p.state.isCompleted) break;
          }
    
          // Load voice
          await RunAnywhere.loadTTSVoice(voiceId);
          setState(() => _selectedVoice = voiceId);
    
          // Preview voice
          await _previewVoice();
        } finally {
          setState(() => _isLoading = false);
        }
      }
    
      Future<void> _previewVoice() async {
        final result = await RunAnywhere.synthesize(
          'Hello, this is a preview of my voice.',
        );
        // Play audio...
      }
    
      @override
      Widget build(BuildContext context) {
        return Column(
          crossAxisAlignment: CrossAxisAlignment.start,
          children: [\
            Text('Select Voice', style: TextStyle(fontWeight: FontWeight.bold)),\
            SizedBox(height: 8),\
            ...voices.map((voice) => RadioListTile<String>(\
              title: Text(voice['name']!),\
              value: voice['id']!,\
              groupValue: _selectedVoice,\
              onChanged: _isLoading ? null : (id) => _selectVoice(id!),\
            )),\
            if (_isLoading) LinearProgressIndicator(),\
          ],
        );
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/tts/voices#system-voice-fallback)

System Voice Fallback
--------------------------------------------------------------------------------------------------

The SDK can fall back to system TTS voices via `flutter_tts` when neural voices aren’t available:

Copy

Ask AI

    // System TTS is automatically available
    // Use it for simple cases or as fallback
    import 'package:flutter_tts/flutter_tts.dart';
    
    final flutterTts = FlutterTts();
    
    // List available system voices
    final voices = await flutterTts.getVoices;
    
    // Speak with system voice
    await flutterTts.speak('Hello world');
    

[​](https://docs.runanywhere.ai/flutter/tts/voices#choosing-between-neural-and-system-voices)

Choosing Between Neural and System Voices
------------------------------------------------------------------------------------------------------------------------------------------

| Feature | Piper (Neural) | System TTS |
| --- | --- | --- |
| Quality | High | Varies |
| Latency | ~100-500ms | ~50ms |
| Size | 20-50MB per voice | 0 (built-in) |
| Offline | Yes | Yes |
| Customization | Limited | More options |

**Recommendation:** Use Piper neural voices for the best quality. Fall back to system TTS for quick responses or when storage is limited.

[​](https://docs.runanywhere.ai/flutter/tts/voices#see-also)

See Also
------------------------------------------------------------------------

[synthesize()\
------------\
\
Basic synthesis](https://docs.runanywhere.ai/flutter/tts/synthesize)
[Voice Agent\
-----------\
\
Complete voice pipeline](https://docs.runanywhere.ai/flutter/voice-agent)

[Streaming TTS](https://docs.runanywhere.ai/flutter/tts/stream)
[Voice Activity Detection](https://docs.runanywhere.ai/flutter/vad)

⌘I

---

# Error Handling - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/error-handling#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Flutter SDK

Error Handling

On this page

*   [Basic Error Handling](https://docs.runanywhere.ai/flutter/error-handling#basic-error-handling)
    
*   [Error Categories](https://docs.runanywhere.ai/flutter/error-handling#error-categories)
    
*   [Handling Specific Errors](https://docs.runanywhere.ai/flutter/error-handling#handling-specific-errors)
    
*   [SDKError Factory Methods](https://docs.runanywhere.ai/flutter/error-handling#sdkerror-factory-methods)
    
*   [STT/TTS Specific Errors](https://docs.runanywhere.ai/flutter/error-handling#stt%2Ftts-specific-errors)
    
*   [Voice Agent Errors](https://docs.runanywhere.ai/flutter/error-handling#voice-agent-errors)
    
*   [Download Errors](https://docs.runanywhere.ai/flutter/error-handling#download-errors)
    
*   [User-Friendly Error Messages](https://docs.runanywhere.ai/flutter/error-handling#user-friendly-error-messages)
    
*   [Error Widget Pattern](https://docs.runanywhere.ai/flutter/error-handling#error-widget-pattern)
    
*   [See Also](https://docs.runanywhere.ai/flutter/error-handling#see-also)
    

The SDK provides comprehensive error handling through `SDKError` with specific error codes for different failure scenarios.

[​](https://docs.runanywhere.ai/flutter/error-handling#basic-error-handling)

Basic Error Handling
----------------------------------------------------------------------------------------------------

Copy

Ask AI

    try {
      await RunAnywhere.loadModel('my-model');
      final result = await RunAnywhere.generate('Hello');
      print(result.text);
    } on SDKError catch (error) {
      print('Error: ${error.message}');
      print('Code: ${error.code}');
    }
    

[​](https://docs.runanywhere.ai/flutter/error-handling#error-categories)

Error Categories
--------------------------------------------------------------------------------------------

| Category | Description | Common Errors |
| --- | --- | --- |
| `general` | General SDK errors | notInitialized |
| `llm` | LLM generation errors | generationFailed |
| `stt` | Speech-to-text errors | sttNotAvailable |
| `tts` | Text-to-speech errors | ttsNotAvailable |
| `voiceAgent` | Voice pipeline errors | voiceAgentNotReady |
| `download` | Model download errors | downloadFailed |
| `validation` | Input validation errors | validationFailed |

[​](https://docs.runanywhere.ai/flutter/error-handling#handling-specific-errors)

Handling Specific Errors
------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    try {
      await RunAnywhere.loadModel('my-model');
      final result = await RunAnywhere.generate('Hello!');
    } on SDKError catch (error) {
      switch (error.code) {
        case SDKErrorCode.notInitialized:
          // SDK not initialized
          showError('Please restart the app');
          await RunAnywhere.initialize();
          break;
    
        case SDKErrorCode.modelNotFound:
          // Model not registered
          showError('Model not found. Please check the model ID.');
          break;
    
        case SDKErrorCode.modelNotDownloaded:
          // Model needs to be downloaded first
          showError('Model not downloaded. Downloading now...');
          await downloadModel('my-model');
          break;
    
        case SDKErrorCode.componentNotReady:
          // Component not loaded
          showError('Model not loaded. Loading now...');
          await RunAnywhere.loadModel('my-model');
          break;
    
        case SDKErrorCode.generationFailed:
          // Generation failed
          showError('Generation failed. Please try again.');
          break;
    
        default:
          showError('Error: ${error.message}');
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/error-handling#sdkerror-factory-methods)

SDKError Factory Methods
------------------------------------------------------------------------------------------------------------

Common error types with factory constructors:

Copy

Ask AI

    // Check error types
    if (error == SDKError.notInitialized()) {
      // Handle not initialized
    }
    
    if (error == SDKError.modelNotFound('model-id')) {
      // Handle model not found
    }
    
    if (error == SDKError.modelNotDownloaded('model-id')) {
      // Handle model not downloaded
    }
    
    if (error == SDKError.componentNotReady('LLM')) {
      // Handle component not ready
    }
    

[​](https://docs.runanywhere.ai/flutter/error-handling#stt/tts-specific-errors)

STT/TTS Specific Errors
----------------------------------------------------------------------------------------------------------

Copy

Ask AI

    try {
      final text = await RunAnywhere.transcribe(audioBytes);
    } on SDKError catch (error) {
      switch (error.code) {
        case SDKErrorCode.sttNotAvailable:
          showError('Speech recognition not available. Load an STT model.');
          break;
        case SDKErrorCode.componentNotReady:
          showError('STT model not loaded.');
          await RunAnywhere.loadSTTModel('whisper-tiny-en');
          break;
        default:
          showError('Transcription failed: ${error.message}');
      }
    }
    
    try {
      final result = await RunAnywhere.synthesize('Hello');
    } on SDKError catch (error) {
      switch (error.code) {
        case SDKErrorCode.ttsNotAvailable:
          showError('Text-to-speech not available. Load a TTS voice.');
          break;
        case SDKErrorCode.componentNotReady:
          showError('TTS voice not loaded.');
          await RunAnywhere.loadTTSVoice('piper-en-us-amy');
          break;
        default:
          showError('Synthesis failed: ${error.message}');
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/error-handling#voice-agent-errors)

Voice Agent Errors
------------------------------------------------------------------------------------------------

Copy

Ask AI

    try {
      final session = await RunAnywhere.startVoiceSession();
    } on SDKError catch (error) {
      if (error.code == SDKErrorCode.voiceAgentNotReady) {
        // Check which component is missing
        final states = RunAnywhere.getVoiceAgentComponentStates();
    
        if (states.stt != ComponentLoadState.loaded) {
          showError('STT model not loaded');
          await RunAnywhere.loadSTTModel('whisper-tiny-en');
        }
        if (states.llm != ComponentLoadState.loaded) {
          showError('LLM model not loaded');
          await RunAnywhere.loadModel('smollm2-360m');
        }
        if (states.tts != ComponentLoadState.loaded) {
          showError('TTS voice not loaded');
          await RunAnywhere.loadTTSVoice('piper-en-us-amy');
        }
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/error-handling#download-errors)

Download Errors
------------------------------------------------------------------------------------------

Copy

Ask AI

    try {
      await for (final progress in RunAnywhere.downloadModel('my-model')) {
        if (progress.state == DownloadProgressState.failed) {
          throw SDKError.downloadFailed('Download failed');
        }
        if (progress.state.isCompleted) break;
      }
    } on SDKError catch (error) {
      showError('Failed to download model: ${error.message}');
      // Suggest checking internet connection
      showRetryOption();
    }
    

[​](https://docs.runanywhere.ai/flutter/error-handling#user-friendly-error-messages)

User-Friendly Error Messages
--------------------------------------------------------------------------------------------------------------------

Create a helper to convert errors to user-friendly messages:

Copy

Ask AI

    String getUserFriendlyMessage(SDKError error) {
      switch (error.code) {
        case SDKErrorCode.notInitialized:
          return 'The app needs to restart. Please close and reopen the app.';
        case SDKErrorCode.modelNotFound:
          return 'The AI model could not be found. Please check your settings.';
        case SDKErrorCode.modelNotDownloaded:
          return 'This feature needs to download data first. Please connect to WiFi.';
        case SDKErrorCode.componentNotReady:
          return 'Loading AI features... Please wait a moment.';
        case SDKErrorCode.generationFailed:
          return 'Something went wrong. Please try again.';
        case SDKErrorCode.sttNotAvailable:
          return 'Voice recognition is not available right now.';
        case SDKErrorCode.ttsNotAvailable:
          return 'Voice output is not available right now.';
        default:
          return 'An error occurred. Please try again later.';
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/error-handling#error-widget-pattern)

Error Widget Pattern
----------------------------------------------------------------------------------------------------

Copy

Ask AI

    class ErrorHandler extends StatelessWidget {
      final Widget child;
      final Function(SDKError) onError;
    
      const ErrorHandler({
        required this.child,
        required this.onError,
      });
    
      @override
      Widget build(BuildContext context) {
        return child;
      }
    
      static void handle(BuildContext context, SDKError error) {
        ScaffoldMessenger.of(context).showSnackBar(
          SnackBar(
            content: Text(getUserFriendlyMessage(error)),
            action: SnackBarAction(
              label: 'Retry',
              onPressed: () {
                // Implement retry logic
              },
            ),
          ),
        );
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/error-handling#see-also)

See Also
----------------------------------------------------------------------------

[Configuration\
-------------\
\
SDK configuration](https://docs.runanywhere.ai/flutter/configuration)
[Best Practices\
--------------\
\
Optimization tips](https://docs.runanywhere.ai/flutter/best-practices)

[Configuration](https://docs.runanywhere.ai/flutter/configuration)
[Best Practices](https://docs.runanywhere.ai/flutter/best-practices)

⌘I

---

# Voice Activity Detection - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/vad#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Flutter SDK

Voice Activity Detection

On this page

*   [Overview](https://docs.runanywhere.ai/flutter/vad#overview)
    
*   [Automatic VAD with Voice Session](https://docs.runanywhere.ai/flutter/vad#automatic-vad-with-voice-session)
    
*   [VoiceSessionConfig for VAD](https://docs.runanywhere.ai/flutter/vad#voicesessionconfig-for-vad)
    
*   [Building a Voice Level Indicator](https://docs.runanywhere.ai/flutter/vad#building-a-voice-level-indicator)
    
*   [Voice Session with VAD Events](https://docs.runanywhere.ai/flutter/vad#voice-session-with-vad-events)
    
*   [VAD Tuning Tips](https://docs.runanywhere.ai/flutter/vad#vad-tuning-tips)
    
*   [See Also](https://docs.runanywhere.ai/flutter/vad#see-also)
    

Voice Activity Detection (VAD) automatically detects speech in audio streams, enabling hands-free voice interfaces.

[​](https://docs.runanywhere.ai/flutter/vad#overview)

Overview
-----------------------------------------------------------------

VAD determines when a user is speaking vs. silent, which is essential for:

*   Knowing when to start/stop recording
*   Triggering transcription at the right time
*   Building push-to-talk or hands-free interfaces

The SDK provides energy-based VAD with Silero VAD for accurate speech detection.

[​](https://docs.runanywhere.ai/flutter/vad#automatic-vad-with-voice-session)

Automatic VAD with Voice Session
-----------------------------------------------------------------------------------------------------------------

The easiest way to use VAD is through the Voice Agent pipeline:

Copy

Ask AI

    final session = await RunAnywhere.startVoiceSession(
      config: VoiceSessionConfig(
        speechThreshold: 0.03,    // Audio level threshold
        silenceDuration: 1.5,     // Seconds of silence before processing
      ),
    );
    
    session.events.listen((event) {
      switch (event) {
        case VoiceSessionListening(:final audioLevel):
          // Show audio level meter
          updateAudioMeter(audioLevel);
        case VoiceSessionSpeechStarted():
          print('User started speaking');
        case VoiceSessionProcessing():
          print('Processing speech...');
      }
    });
    

[​](https://docs.runanywhere.ai/flutter/vad#voicesessionconfig-for-vad)

VoiceSessionConfig for VAD
-----------------------------------------------------------------------------------------------------

| Parameter | Type | Default | Description |
| --- | --- | --- | --- |
| `speechThreshold` | `double` | 0.03 | Audio level to trigger speech detection |
| `silenceDuration` | `double` | 1.5 | Seconds of silence before processing |

Copy

Ask AI

    // More sensitive (quieter environments)
    final quietConfig = VoiceSessionConfig(
      speechThreshold: 0.01,
      silenceDuration: 1.0,
    );
    
    // Less sensitive (noisy environments)
    final noisyConfig = VoiceSessionConfig(
      speechThreshold: 0.1,
      silenceDuration: 2.0,
    );
    

[​](https://docs.runanywhere.ai/flutter/vad#building-a-voice-level-indicator)

Building a Voice Level Indicator
-----------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class VoiceLevelIndicator extends StatefulWidget {
      final Stream<double> audioLevelStream;
    
      const VoiceLevelIndicator({required this.audioLevelStream});
    
      @override
      _VoiceLevelIndicatorState createState() => _VoiceLevelIndicatorState();
    }
    
    class _VoiceLevelIndicatorState extends State<VoiceLevelIndicator> {
      double _level = 0.0;
      StreamSubscription? _subscription;
    
      @override
      void initState() {
        super.initState();
        _subscription = widget.audioLevelStream.listen((level) {
          setState(() => _level = level);
        });
      }
    
      @override
      Widget build(BuildContext context) {
        return Container(
          height: 10,
          child: LinearProgressIndicator(
            value: _level.clamp(0.0, 1.0),
            backgroundColor: Colors.grey[300],
            valueColor: AlwaysStoppedAnimation(
              _level > 0.03 ? Colors.green : Colors.grey,
            ),
          ),
        );
      }
    
      @override
      void dispose() {
        _subscription?.cancel();
        super.dispose();
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/vad#voice-session-with-vad-events)

Voice Session with VAD Events
-----------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class VoiceAssistant extends StatefulWidget {
      @override
      _VoiceAssistantState createState() => _VoiceAssistantState();
    }
    
    class _VoiceAssistantState extends State<VoiceAssistant> {
      VoiceSessionHandle? _session;
      double _audioLevel = 0.0;
      bool _isSpeaking = false;
      String _status = 'Ready';
    
      Future<void> _startSession() async {
        _session = await RunAnywhere.startVoiceSession(
          config: VoiceSessionConfig(
            speechThreshold: 0.03,
            silenceDuration: 1.5,
            autoPlayTTS: true,
            continuousMode: true,
          ),
        );
    
        _session!.events.listen((event) {
          setState(() {
            switch (event) {
              case VoiceSessionListening(:final audioLevel):
                _audioLevel = audioLevel;
                _status = 'Listening...';
              case VoiceSessionSpeechStarted():
                _isSpeaking = true;
                _status = 'Speaking detected';
              case VoiceSessionProcessing():
                _isSpeaking = false;
                _status = 'Processing...';
              case VoiceSessionTranscribed(:final text):
                _status = 'You: $text';
              case VoiceSessionResponded(:final text):
                _status = 'AI: $text';
              case VoiceSessionSpeaking():
                _status = 'AI speaking...';
              case VoiceSessionTurnCompleted():
                _status = 'Ready';
              case VoiceSessionStopped():
                _status = 'Stopped';
            }
          });
        });
      }
    
      void _stopSession() {
        _session?.stop();
      }
    
      @override
      Widget build(BuildContext context) {
        return Column(
          mainAxisAlignment: MainAxisAlignment.center,
          children: [\
            // Audio level indicator\
            Container(\
              width: 100,\
              height: 100,\
              decoration: BoxDecoration(\
                shape: BoxShape.circle,\
                color: _isSpeaking ? Colors.green : Colors.blue,\
                boxShadow: [\
                  BoxShadow(\
                    color: (_isSpeaking ? Colors.green : Colors.blue)\
                        .withOpacity(0.5),\
                    blurRadius: _audioLevel * 50,\
                    spreadRadius: _audioLevel * 20,\
                  ),\
                ],\
              ),\
              child: Icon(\
                _isSpeaking ? Icons.mic : Icons.mic_none,\
                color: Colors.white,\
                size: 48,\
              ),\
            ),\
    \
            SizedBox(height: 24),\
            Text(_status, textAlign: TextAlign.center),\
    \
            SizedBox(height: 24),\
            ElevatedButton(\
              onPressed: _session == null ? _startSession : _stopSession,\
              child: Text(_session == null ? 'Start' : 'Stop'),\
            ),\
          ],
        );
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/vad#vad-tuning-tips)

VAD Tuning Tips
-------------------------------------------------------------------------------

Threshold Adjustment

Lower thresholds (0.01) detect quieter speech but may trigger on background noise. Higher thresholds (0.1) require louder speech but are more robust to noise.

Silence Duration

Shorter duration (0.5s) feels more responsive but may cut off pauses mid-sentence. Longer duration (2.0s) allows natural pauses but feels slower.

Environment Calibration

Consider measuring ambient noise level and adjusting threshold dynamically.

[​](https://docs.runanywhere.ai/flutter/vad#see-also)

See Also
-----------------------------------------------------------------

[Voice Agent\
-----------\
\
Complete voice pipeline](https://docs.runanywhere.ai/flutter/voice-agent)
[Speech-to-Text\
--------------\
\
Audio transcription](https://docs.runanywhere.ai/flutter/stt/transcribe)

[Voices](https://docs.runanywhere.ai/flutter/tts/voices)
[Voice Agent](https://docs.runanywhere.ai/flutter/voice-agent)

⌘I

---

# STT Options - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/stt/options#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

STT Options

On this page

*   [Model Selection](https://docs.runanywhere.ai/flutter/stt/options#model-selection)
    
*   [Register Multiple Models](https://docs.runanywhere.ai/flutter/stt/options#register-multiple-models)
    
*   [Switch Models at Runtime](https://docs.runanywhere.ai/flutter/stt/options#switch-models-at-runtime)
    
*   [Memory Management](https://docs.runanywhere.ai/flutter/stt/options#memory-management)
    
*   [Audio Preprocessing Tips](https://docs.runanywhere.ai/flutter/stt/options#audio-preprocessing-tips)
    
*   [Error Handling](https://docs.runanywhere.ai/flutter/stt/options#error-handling)
    
*   [Best Practices](https://docs.runanywhere.ai/flutter/stt/options#best-practices)
    
*   [See Also](https://docs.runanywhere.ai/flutter/stt/options#see-also)
    

Configure STT behavior for different use cases and languages.

[​](https://docs.runanywhere.ai/flutter/stt/options#model-selection)

Model Selection
---------------------------------------------------------------------------------------

Choose the right model based on your needs:

| Model | Size | Languages | Quality | Speed | Use Case |
| --- | --- | --- | --- | --- | --- |
| `whisper-tiny.en` | ~75MB | English | Good | Fastest | Quick commands |
| `whisper-base.en` | ~150MB | English | Better | Fast | General use |
| `whisper-small.en` | ~250MB | English | Best | Medium | Accuracy-critical |
| `whisper-tiny` | ~75MB | Multi | Good | Fast | Multilingual apps |

[​](https://docs.runanywhere.ai/flutter/stt/options#register-multiple-models)

Register Multiple Models
---------------------------------------------------------------------------------------------------------

Copy

Ask AI

    // English-optimized for primary use
    Onnx.addModel(
      id: 'whisper-tiny-en',
      name: 'Whisper Tiny English',
      url: 'https://github.com/RunanywhereAI/sherpa-onnx/releases/download/runanywhere-models-v1/sherpa-onnx-whisper-tiny.en.tar.gz',
      modality: ModelCategory.speechRecognition,
    );
    
    // Multilingual for international users
    Onnx.addModel(
      id: 'whisper-tiny-multi',
      name: 'Whisper Tiny Multilingual',
      url: 'https://github.com/RunanywhereAI/sherpa-onnx/releases/download/runanywhere-models-v1/sherpa-onnx-whisper-tiny.tar.gz',
      modality: ModelCategory.speechRecognition,
    );
    

[​](https://docs.runanywhere.ai/flutter/stt/options#switch-models-at-runtime)

Switch Models at Runtime
---------------------------------------------------------------------------------------------------------

Copy

Ask AI

    // Load English model by default
    await RunAnywhere.loadSTTModel('whisper-tiny-en');
    
    // Switch to multilingual when needed
    Future<void> switchToMultilingual() async {
      await RunAnywhere.unloadSTTModel();
      await RunAnywhere.loadSTTModel('whisper-tiny-multi');
    }
    

[​](https://docs.runanywhere.ai/flutter/stt/options#memory-management)

Memory Management
-------------------------------------------------------------------------------------------

Unload STT model when not needed to free memory:

Copy

Ask AI

    // Check if loaded
    if (RunAnywhere.isSTTModelLoaded) {
      print('Current STT model: ${RunAnywhere.currentSTTModelId}');
    }
    
    // Unload to free memory
    await RunAnywhere.unloadSTTModel();
    

[​](https://docs.runanywhere.ai/flutter/stt/options#audio-preprocessing-tips)

Audio Preprocessing Tips
---------------------------------------------------------------------------------------------------------

Sample Rate Conversion

If your audio isn’t 16kHz, convert it before transcription: `dart // Example: Convert 44.1kHz to 16kHz // Use a package like 'flutter_sound' for resampling`

Noise Reduction

For noisy environments, consider preprocessing audio: - Apply a high-pass filter to remove low-frequency noise - Normalize audio levels - Remove silence at beginning/end

Audio Format

Always ensure correct format: - PCM16 (16-bit signed integer) - 16,000 Hz sample rate - Mono (single channel)

[​](https://docs.runanywhere.ai/flutter/stt/options#error-handling)

Error Handling
-------------------------------------------------------------------------------------

Copy

Ask AI

    try {
      final text = await RunAnywhere.transcribe(audioBytes);
      print('Transcribed: $text');
    } on SDKError catch (e) {
      switch (e.code) {
        case SDKErrorCode.sttNotAvailable:
          print('STT not available. Load an STT model first.');
          break;
        case SDKErrorCode.componentNotReady:
          print('STT model not loaded.');
          break;
        default:
          print('STT error: ${e.message}');
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/stt/options#best-practices)

Best Practices
-------------------------------------------------------------------------------------

**Start with the smallest model** that meets your accuracy needs. You can always upgrade later if needed.

1.  **Preload during idle time** — Download and load STT model before user needs it
2.  **Use English-specific models** — They’re smaller and more accurate for English
3.  **Handle empty audio** — Check audio length before transcribing
4.  **Provide feedback** — Show transcription progress to users

[​](https://docs.runanywhere.ai/flutter/stt/options#see-also)

See Also
-------------------------------------------------------------------------

[transcribe()\
------------\
\
Basic transcription](https://docs.runanywhere.ai/flutter/stt/transcribe)
[Voice Agent\
-----------\
\
Complete voice pipeline](https://docs.runanywhere.ai/flutter/voice-agent)

[Streaming STT](https://docs.runanywhere.ai/flutter/stt/stream)
[synthesize()](https://docs.runanywhere.ai/flutter/tts/synthesize)

⌘I

---

# Installation - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/installation#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Kotlin SDK

Installation

On this page

*   [Gradle Setup](https://docs.runanywhere.ai/kotlin/installation#gradle-setup)
    
*   [Module Overview](https://docs.runanywhere.ai/kotlin/installation#module-overview)
    
*   [Supported Model Formats](https://docs.runanywhere.ai/kotlin/installation#supported-model-formats)
    
*   [Android Manifest](https://docs.runanywhere.ai/kotlin/installation#android-manifest)
    
*   [ProGuard Rules](https://docs.runanywhere.ai/kotlin/installation#proguard-rules)
    
*   [Next Steps](https://docs.runanywhere.ai/kotlin/installation#next-steps)
    

[​](https://docs.runanywhere.ai/kotlin/installation#gradle-setup)

Gradle Setup
---------------------------------------------------------------------------------

Add the RunAnywhere SDK to your module-level `build.gradle.kts`:

build.gradle.kts (Module: app)

Copy

Ask AI

    dependencies {
        // Core SDK (required)
        implementation("com.runanywhere.sdk:runanywhere-kotlin:0.1.4")
    
        // LLM support - llama.cpp backend (~34MB)
        implementation("com.runanywhere.sdk:runanywhere-core-llamacpp:0.1.4")
    
        // STT/TTS/VAD support - ONNX backend (~25MB)
        implementation("com.runanywhere.sdk:runanywhere-core-onnx:0.1.4")
    }
    

Include only the modules you need. Each native module adds to your APK size: - **runanywhere-core-llamacpp**: ~34MB (for LLM text generation) - **runanywhere-core-onnx**: ~25MB (for STT, TTS, and VAD)

[​](https://docs.runanywhere.ai/kotlin/installation#module-overview)

Module Overview
---------------------------------------------------------------------------------------

| Module | Size | Features |
| --- | --- | --- |
| `runanywhere-kotlin` | ~2MB | Core SDK, model management, events |
| `runanywhere-core-llamacpp` | ~34MB | LLM text generation (GGUF models) |
| `runanywhere-core-onnx` | ~25MB | STT, TTS, VAD (ONNX models) |

[​](https://docs.runanywhere.ai/kotlin/installation#supported-model-formats)

Supported Model Formats
-------------------------------------------------------------------------------------------------------

| Format | Extension | Backend | Use Case |
| --- | --- | --- | --- |
| GGUF | `.gguf` | llama.cpp | LLM text generation |
| ONNX | `.onnx` | ONNX Runtime | STT, TTS, VAD |
| ORT | `.ort` | ONNX Runtime | Optimized STT/TTS |

[​](https://docs.runanywhere.ai/kotlin/installation#android-manifest)

Android Manifest
-----------------------------------------------------------------------------------------

Add the required permissions to your `AndroidManifest.xml`:

AndroidManifest.xml

Copy

Ask AI

    <uses-permission android:name="android.permission.INTERNET" />
    <uses-permission android:name="android.permission.RECORD_AUDIO" />
    

`INTERNET` is required for model downloads. `RECORD_AUDIO` is only needed if you’re using STT or VAD with microphone input.

[​](https://docs.runanywhere.ai/kotlin/installation#proguard-rules)

ProGuard Rules
-------------------------------------------------------------------------------------

If using ProGuard/R8 minification, add these rules:

proguard-rules.pro

Copy

Ask AI

    -keep class com.runanywhere.sdk.** { *; }
    -keepclassmembers class com.runanywhere.sdk.** { *; }
    

[​](https://docs.runanywhere.ai/kotlin/installation#next-steps)

Next Steps
-----------------------------------------------------------------------------

[Quick Start\
-----------\
\
Initialize the SDK and run your first inference →](https://docs.runanywhere.ai/kotlin/quick-start)

[Introduction](https://docs.runanywhere.ai/kotlin/introduction)
[Quick Start](https://docs.runanywhere.ai/kotlin/quick-start)

⌘I

---

# System Prompts - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/llm/system-prompts#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

System Prompts

On this page

*   [Basic Usage](https://docs.runanywhere.ai/flutter/llm/system-prompts#basic-usage)
    
*   [Common Patterns](https://docs.runanywhere.ai/flutter/llm/system-prompts#common-patterns)
    
*   [Coding Assistant](https://docs.runanywhere.ai/flutter/llm/system-prompts#coding-assistant)
    
*   [Customer Support Bot](https://docs.runanywhere.ai/flutter/llm/system-prompts#customer-support-bot)
    
*   [JSON Output](https://docs.runanywhere.ai/flutter/llm/system-prompts#json-output)
    
*   [Character/Persona](https://docs.runanywhere.ai/flutter/llm/system-prompts#character%2Fpersona)
    
*   [Tips for Effective System Prompts](https://docs.runanywhere.ai/flutter/llm/system-prompts#tips-for-effective-system-prompts)
    
*   [Reusable Prompt Class](https://docs.runanywhere.ai/flutter/llm/system-prompts#reusable-prompt-class)
    
*   [See Also](https://docs.runanywhere.ai/flutter/llm/system-prompts#see-also)
    

System prompts set the context and behavior for your AI model. Use them to create specialized assistants, enforce response formats, or constrain the model’s persona.

[​](https://docs.runanywhere.ai/flutter/llm/system-prompts#basic-usage)

Basic Usage
--------------------------------------------------------------------------------------

Copy

Ask AI

    final result = await RunAnywhere.generate(
      'Write a function to sort an array',
      options: LLMGenerationOptions(
        maxTokens: 300,
        systemPrompt: 'You are a helpful coding assistant. Write clean, well-commented code.',
      ),
    );
    

[​](https://docs.runanywhere.ai/flutter/llm/system-prompts#common-patterns)

Common Patterns
----------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/flutter/llm/system-prompts#coding-assistant)

Coding Assistant

Copy

Ask AI

    const codingPrompt = '''
    You are an expert programmer. Follow these rules:
    - Write clean, readable code with comments
    - Use best practices and design patterns
    - Explain your approach briefly before the code
    - Handle edge cases
    ''';
    
    final result = await RunAnywhere.generate(
      'Implement a binary search in Dart',
      options: LLMGenerationOptions(
        systemPrompt: codingPrompt,
        maxTokens: 400,
      ),
    );
    

### 

[​](https://docs.runanywhere.ai/flutter/llm/system-prompts#customer-support-bot)

Customer Support Bot

Copy

Ask AI

    const supportPrompt = '''
    You are a friendly customer support agent for TechCorp.
    - Be helpful and empathetic
    - Keep responses concise (2-3 sentences)
    - If you don't know something, say so
    - Never make up information about products
    ''';
    

### 

[​](https://docs.runanywhere.ai/flutter/llm/system-prompts#json-output)

JSON Output

Copy

Ask AI

    const jsonPrompt = '''
    You are a data extraction assistant.
    Always respond with valid JSON only, no additional text.
    Extract the requested information into the specified format.
    ''';
    
    final result = await RunAnywhere.generate(
      'Extract: "John Smith, 25 years old, engineer at Google"',
      options: LLMGenerationOptions(
        systemPrompt: jsonPrompt,
        maxTokens: 100,
        temperature: 0.1,  // Lower temperature for consistent format
      ),
    );
    
    // Parse the JSON response
    final data = jsonDecode(result.text);
    

### 

[​](https://docs.runanywhere.ai/flutter/llm/system-prompts#character/persona)

Character/Persona

Copy

Ask AI

    const personaPrompt = '''
    You are Captain Nova, a space explorer from the year 3000.
    - Speak with enthusiasm about space and technology
    - Reference your adventures across the galaxy
    - Use occasional space-themed expressions
    - Stay in character at all times
    ''';
    

[​](https://docs.runanywhere.ai/flutter/llm/system-prompts#tips-for-effective-system-prompts)

Tips for Effective System Prompts
----------------------------------------------------------------------------------------------------------------------------------

Be Specific

Vague prompts lead to inconsistent results. Specify exactly what you want: ❌ “Be helpful” ✅ “Provide step-by-step instructions. Use bullet points. Keep each step under 20 words.”

Use Examples

Show the model what you expect: `Format responses like this example: Q: What is 2+2? A: **4** — Two plus two equals four.`

Set Boundaries

Tell the model what NOT to do: `- Do not provide medical advice - Do not generate harmful content - If asked about topics outside your expertise, politely decline`

Keep It Concise

Shorter prompts use fewer tokens and leave more room for generation: ❌ Long, rambling instructions with repetition ✅ Clear, bullet-pointed rules

[​](https://docs.runanywhere.ai/flutter/llm/system-prompts#reusable-prompt-class)

Reusable Prompt Class
----------------------------------------------------------------------------------------------------------

Create a helper class for managing prompts:

Copy

Ask AI

    class SystemPrompts {
      static const String coding = '''
    You are an expert programmer. Write clean, well-documented code.
    ''';
    
      static const String creative = '''
    You are a creative writing assistant. Be imaginative and engaging.
    ''';
    
      static String forDomain(String domain) => '''
    You are an expert in $domain. Provide accurate, helpful information.
    ''';
    }
    
    // Usage
    final result = await RunAnywhere.generate(
      prompt,
      options: LLMGenerationOptions(
        systemPrompt: SystemPrompts.coding,
      ),
    );
    

[​](https://docs.runanywhere.ai/flutter/llm/system-prompts#see-also)

See Also
--------------------------------------------------------------------------------

[generate()\
----------\
\
Full generation API](https://docs.runanywhere.ai/flutter/llm/generate)
[Best Practices\
--------------\
\
Optimization tips](https://docs.runanywhere.ai/flutter/best-practices)

[generateStream()](https://docs.runanywhere.ai/flutter/llm/stream)
[transcribe()](https://docs.runanywhere.ai/flutter/stt/transcribe)

⌘I

---

# synthesize() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/tts/synthesize#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

synthesize()

On this page

*   [Basic Usage](https://docs.runanywhere.ai/flutter/tts/synthesize#basic-usage)
    
*   [TTSResult](https://docs.runanywhere.ai/flutter/tts/synthesize#ttsresult)
    
*   [Parameters](https://docs.runanywhere.ai/flutter/tts/synthesize#parameters)
    
*   [Setup](https://docs.runanywhere.ai/flutter/tts/synthesize#setup)
    
*   [1\. Register ONNX Backend](https://docs.runanywhere.ai/flutter/tts/synthesize#1-register-onnx-backend)
    
*   [2\. Add TTS Voice](https://docs.runanywhere.ai/flutter/tts/synthesize#2-add-tts-voice)
    
*   [3\. Download & Load](https://docs.runanywhere.ai/flutter/tts/synthesize#3-download-%26-load)
    
*   [Playing Audio](https://docs.runanywhere.ai/flutter/tts/synthesize#playing-audio)
    
*   [Complete Example](https://docs.runanywhere.ai/flutter/tts/synthesize#complete-example)
    
*   [See Also](https://docs.runanywhere.ai/flutter/tts/synthesize#see-also)
    

Synthesize natural-sounding speech from text using on-device Piper TTS models.

[​](https://docs.runanywhere.ai/flutter/tts/synthesize#basic-usage)

Basic Usage
----------------------------------------------------------------------------------

Copy

Ask AI

    // Load a TTS voice
    await RunAnywhere.loadTTSVoice('vits-piper-en_US-lessac-medium');
    
    // Synthesize speech
    final result = await RunAnywhere.synthesize(
      'Hello! Welcome to RunAnywhere.',
      rate: 1.0,
      pitch: 1.0,
    );
    
    // result.samples contains PCM Float32 audio
    // result.sampleRate is typically 22050 Hz
    print('Duration: ${result.durationSeconds}s');
    

[​](https://docs.runanywhere.ai/flutter/tts/synthesize#ttsresult)

TTSResult
------------------------------------------------------------------------------

| Property | Type | Description |
| --- | --- | --- |
| `samples` | `Float32List` | Audio samples (PCM float) |
| `sampleRate` | `int` | Sample rate in Hz (typically 22050) |
| `durationMs` | `int` | Duration in milliseconds |
| `durationSeconds` | `double` | Duration in seconds |
| `numSamples` | `int` | Number of audio samples |

[​](https://docs.runanywhere.ai/flutter/tts/synthesize#parameters)

Parameters
--------------------------------------------------------------------------------

| Parameter | Type | Default | Description |
| --- | --- | --- | --- |
| `rate` | `double` | 1.0 | Speech rate (0.5 to 2.0) |
| `pitch` | `double` | 1.0 | Voice pitch (0.5 to 2.0) |
| `volume` | `double` | 1.0 | Volume (0.0 to 1.0) |

Copy

Ask AI

    // Slower, lower pitch for emphasis
    final result = await RunAnywhere.synthesize(
      'Important announcement!',
      rate: 0.8,
      pitch: 0.9,
      volume: 1.0,
    );
    

[​](https://docs.runanywhere.ai/flutter/tts/synthesize#setup)

Setup
----------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/flutter/tts/synthesize#1-register-onnx-backend)

1\. Register ONNX Backend

Copy

Ask AI

    import 'package:runanywhere_onnx/runanywhere_onnx.dart';
    
    await Onnx.register();
    

### 

[​](https://docs.runanywhere.ai/flutter/tts/synthesize#2-add-tts-voice)

2\. Add TTS Voice

Copy

Ask AI

    Onnx.addModel(
      id: 'vits-piper-en_US-lessac-medium',
      name: 'Piper US English (Lessac)',
      url: 'https://github.com/RunanywhereAI/sherpa-onnx/releases/download/runanywhere-models-v1/vits-piper-en_US-lessac-medium.tar.bz2',
      modality: ModelCategory.textToSpeech,
    );
    

### 

[​](https://docs.runanywhere.ai/flutter/tts/synthesize#3-download-&-load)

3\. Download & Load

Copy

Ask AI

    // Download
    await for (final progress in RunAnywhere.downloadModel('vits-piper-en_US-lessac-medium')) {
      print('${(progress.percentage * 100).toStringAsFixed(1)}%');
      if (progress.state.isCompleted) break;
    }
    
    // Load
    await RunAnywhere.loadTTSVoice('vits-piper-en_US-lessac-medium');
    

[​](https://docs.runanywhere.ai/flutter/tts/synthesize#playing-audio)

Playing Audio
--------------------------------------------------------------------------------------

Use a package like `audioplayers` to play the synthesized audio:

Copy

Ask AI

    import 'package:audioplayers/audioplayers.dart';
    
    final player = AudioPlayer();
    
    Future<void> speakText(String text) async {
      final result = await RunAnywhere.synthesize(text);
    
      // Convert Float32 samples to bytes for playback
      final bytes = _convertToBytes(result.samples, result.sampleRate);
    
      // Play using audioplayers
      await player.play(BytesSource(bytes));
    }
    
    Uint8List _convertToBytes(Float32List samples, int sampleRate) {
      // Convert Float32 PCM to Int16 PCM WAV format
      final int16Samples = Int16List(samples.length);
      for (var i = 0; i < samples.length; i++) {
        int16Samples[i] = (samples[i] * 32767).clamp(-32768, 32767).toInt();
      }
    
      // Create WAV header + data
      return _createWavFile(int16Samples, sampleRate);
    }
    

[​](https://docs.runanywhere.ai/flutter/tts/synthesize#complete-example)

Complete Example
--------------------------------------------------------------------------------------------

Copy

Ask AI

    class TextToSpeechDemo extends StatefulWidget {
      @override
      _TextToSpeechDemoState createState() => _TextToSpeechDemoState();
    }
    
    class _TextToSpeechDemoState extends State<TextToSpeechDemo> {
      final _controller = TextEditingController();
      final _player = AudioPlayer();
      bool _isSpeaking = false;
      bool _isLoaded = false;
    
      @override
      void initState() {
        super.initState();
        _initTTS();
      }
    
      Future<void> _initTTS() async {
        // Download if needed
        final models = await RunAnywhere.availableModels();
        final ttsModel = models.firstWhere(
          (m) => m.id == 'vits-piper-en_US-lessac-medium',
        );
    
        if (!ttsModel.isDownloaded) {
          await for (final p in RunAnywhere.downloadModel(ttsModel.id)) {
            if (p.state.isCompleted) break;
          }
        }
    
        // Load voice
        await RunAnywhere.loadTTSVoice('vits-piper-en_US-lessac-medium');
        setState(() => _isLoaded = true);
      }
    
      Future<void> _speak() async {
        if (_controller.text.isEmpty) return;
    
        setState(() => _isSpeaking = true);
    
        try {
          final result = await RunAnywhere.synthesize(_controller.text);
          // Play audio (implementation depends on your audio setup)
          await _playAudio(result);
        } finally {
          setState(() => _isSpeaking = false);
        }
      }
    
      @override
      Widget build(BuildContext context) {
        return Column(
          children: [\
            TextField(\
              controller: _controller,\
              decoration: InputDecoration(\
                hintText: 'Enter text to speak...',\
              ),\
              maxLines: 3,\
            ),\
            SizedBox(height: 16),\
            ElevatedButton.icon(\
              onPressed: _isLoaded && !_isSpeaking ? _speak : null,\
              icon: Icon(_isSpeaking ? Icons.stop : Icons.volume_up),\
              label: Text(_isSpeaking ? 'Speaking...' : 'Speak'),\
            ),\
          ],
        );
      }
    
      @override
      void dispose() {
        _controller.dispose();
        _player.dispose();
        super.dispose();
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/tts/synthesize#see-also)

See Also
----------------------------------------------------------------------------

[Streaming TTS\
-------------\
\
Stream audio as it’s generated](https://docs.runanywhere.ai/flutter/tts/stream)
[Available Voices\
----------------\
\
Browse voice options](https://docs.runanywhere.ai/flutter/tts/voices)

[STT Options](https://docs.runanywhere.ai/flutter/stt/options)
[Streaming TTS](https://docs.runanywhere.ai/flutter/tts/stream)

⌘I

---

# Quick Start - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/quick-start#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Flutter SDK

Quick Start

On this page

*   [Complete Example](https://docs.runanywhere.ai/flutter/quick-start#complete-example)
    
*   [Step-by-Step Breakdown](https://docs.runanywhere.ai/flutter/quick-start#step-by-step-breakdown)
    
*   [1\. Initialize the SDK](https://docs.runanywhere.ai/flutter/quick-start#1-initialize-the-sdk)
    
*   [2\. Register Backend Modules](https://docs.runanywhere.ai/flutter/quick-start#2-register-backend-modules)
    
*   [3\. Register Models](https://docs.runanywhere.ai/flutter/quick-start#3-register-models)
    
*   [4\. Download & Load Models](https://docs.runanywhere.ai/flutter/quick-start#4-download-%26-load-models)
    
*   [5\. Generate Text](https://docs.runanywhere.ai/flutter/quick-start#5-generate-text)
    
*   [Environment Options](https://docs.runanywhere.ai/flutter/quick-start#environment-options)
    
*   [Starter Example Repository](https://docs.runanywhere.ai/flutter/quick-start#starter-example-repository)
    
*   [Next Steps](https://docs.runanywhere.ai/flutter/quick-start#next-steps)
    

**Want a complete working app?** Clone our [Flutter Starter Example](https://github.com/RunanywhereAI/flutter-starter-example)
 to get started immediately with a full-featured demo app.

[​](https://docs.runanywhere.ai/flutter/quick-start#complete-example)

Complete Example
-----------------------------------------------------------------------------------------

Here’s a complete example to get you started with on-device text generation:

Copy

Ask AI

    import 'package:flutter/material.dart';
    import 'package:runanywhere/runanywhere.dart';
    import 'package:runanywhere_llamacpp/runanywhere_llamacpp.dart';
    
    void main() async {
      WidgetsFlutterBinding.ensureInitialized();
    
      // 1. Initialize SDK
      await RunAnywhere.initialize();
    
      // 2. Register LlamaCpp backend
      await LlamaCpp.register();
    
      // 3. Add a model
      LlamaCpp.addModel(
        id: 'smollm2-360m-q8_0',
        name: 'SmolLM2 360M Q8_0',
        url: 'https://huggingface.co/prithivMLmods/SmolLM2-360M-GGUF/resolve/main/SmolLM2-360M.Q8_0.gguf',
        memoryRequirement: 500000000,
      );
    
      runApp(const MyApp());
    }
    
    class MyApp extends StatelessWidget {
      const MyApp({super.key});
    
      @override
      Widget build(BuildContext context) {
        return MaterialApp(
          home: ChatScreen(),
        );
      }
    }
    
    class ChatScreen extends StatefulWidget {
      @override
      _ChatScreenState createState() => _ChatScreenState();
    }
    
    class _ChatScreenState extends State<ChatScreen> {
      String _response = '';
      bool _isLoading = false;
      bool _isModelLoaded = false;
    
      Future<void> _loadModel() async {
        setState(() => _isLoading = true);
    
        try {
          // Download model (with progress)
          await for (final progress in RunAnywhere.downloadModel('smollm2-360m-q8_0')) {
            print('Download: ${(progress.percentage * 100).toStringAsFixed(1)}%');
            if (progress.state.isCompleted) break;
          }
    
          // Load model
          await RunAnywhere.loadModel('smollm2-360m-q8_0');
          setState(() => _isModelLoaded = true);
        } catch (e) {
          setState(() => _response = 'Error: $e');
        } finally {
          setState(() => _isLoading = false);
        }
      }
    
      Future<void> _generateText() async {
        setState(() => _isLoading = true);
    
        try {
          final result = await RunAnywhere.generate(
            'Explain quantum computing in simple terms',
            options: LLMGenerationOptions(
              maxTokens: 200,
              temperature: 0.7,
            ),
          );
    
          setState(() => _response = result.text);
          print('Speed: ${result.tokensPerSecond.toStringAsFixed(1)} tok/s');
        } catch (e) {
          setState(() => _response = 'Error: $e');
        } finally {
          setState(() => _isLoading = false);
        }
      }
    
      @override
      Widget build(BuildContext context) {
        return Scaffold(
          appBar: AppBar(title: Text('RunAnywhere Demo')),
          body: Padding(
            padding: EdgeInsets.all(16),
            child: Column(
              children: [\
                if (!_isModelLoaded)\
                  ElevatedButton(\
                    onPressed: _isLoading ? null : _loadModel,\
                    child: Text(_isLoading ? 'Loading...' : 'Download & Load Model'),\
                  ),\
                if (_isModelLoaded)\
                  ElevatedButton(\
                    onPressed: _isLoading ? null : _generateText,\
                    child: Text(_isLoading ? 'Generating...' : 'Generate'),\
                  ),\
                SizedBox(height: 20),\
                Expanded(\
                  child: SingleChildScrollView(\
                    child: Text(_response),\
                  ),\
                ),\
              ],
            ),
          ),
        );
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/quick-start#step-by-step-breakdown)

Step-by-Step Breakdown
-----------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/flutter/quick-start#1-initialize-the-sdk)

1\. Initialize the SDK

Copy

Ask AI

    await RunAnywhere.initialize();
    

Initialize once at app startup. Development mode requires no API key.

### 

[​](https://docs.runanywhere.ai/flutter/quick-start#2-register-backend-modules)

2\. Register Backend Modules

Copy

Ask AI

    await LlamaCpp.register();  // For LLM
    await Onnx.register();      // For STT/TTS (optional)
    

Register the backends you need. Each backend is a separate package.

### 

[​](https://docs.runanywhere.ai/flutter/quick-start#3-register-models)

3\. Register Models

Copy

Ask AI

    LlamaCpp.addModel(
      id: 'smollm2-360m-q8_0',
      name: 'SmolLM2 360M Q8_0',
      url: 'https://huggingface.co/.../SmolLM2-360M.Q8_0.gguf',
      memoryRequirement: 500000000,
    );
    

Register models with their download URLs. Models are downloaded on-demand.

### 

[​](https://docs.runanywhere.ai/flutter/quick-start#4-download-&-load-models)

4\. Download & Load Models

Copy

Ask AI

    // Download with progress tracking
    await for (final progress in RunAnywhere.downloadModel('smollm2-360m-q8_0')) {
      print('${(progress.percentage * 100).toStringAsFixed(1)}%');
      if (progress.state.isCompleted) break;
    }
    
    // Load into memory
    await RunAnywhere.loadModel('smollm2-360m-q8_0');
    

### 

[​](https://docs.runanywhere.ai/flutter/quick-start#5-generate-text)

5\. Generate Text

Copy

Ask AI

    // Simple chat
    final response = await RunAnywhere.chat('Hello!');
    
    // Full generation with metrics
    final result = await RunAnywhere.generate(
      'Your prompt here',
      options: LLMGenerationOptions(maxTokens: 200),
    );
    print('Response: ${result.text}');
    print('Speed: ${result.tokensPerSecond} tok/s');
    

[​](https://docs.runanywhere.ai/flutter/quick-start#environment-options)

Environment Options
-----------------------------------------------------------------------------------------------

| Environment | Description | API Key Required |
| --- | --- | --- |
| `.development` | Verbose logging, local-only | No  |
| `.staging` | Testing with real services | Yes |
| `.production` | Minimal logging, telemetry | Yes |

Copy

Ask AI

    // Production initialization
    await RunAnywhere.initialize(
      apiKey: 'your-api-key',
      baseURL: 'https://api.runanywhere.ai',
      environment: SDKEnvironment.production,
    );
    

[​](https://docs.runanywhere.ai/flutter/quick-start#starter-example-repository)

Starter Example Repository
-------------------------------------------------------------------------------------------------------------

For a complete working application with all features demonstrated, clone our starter example:

Copy

Ask AI

    git clone https://github.com/RunanywhereAI/flutter-starter-example.git
    cd flutter-starter-example
    flutter pub get
    cd ios && pod install && cd ..
    flutter run
    

The [Flutter Starter Example](https://github.com/RunanywhereAI/flutter-starter-example)
 includes:

*   AI Chat with streaming responses
*   Speech-to-Text transcription
*   Text-to-Speech synthesis
*   Voice Assistant pipeline
*   Model download and management UI

[​](https://docs.runanywhere.ai/flutter/quick-start#next-steps)

Next Steps
-----------------------------------------------------------------------------

[LLM Chat\
--------\
\
Simple text generation](https://docs.runanywhere.ai/flutter/llm/chat)
[Streaming\
---------\
\
Real-time token streaming](https://docs.runanywhere.ai/flutter/llm/stream)
[Speech-to-Text\
--------------\
\
Audio transcription](https://docs.runanywhere.ai/flutter/stt/transcribe)
[Text-to-Speech\
--------------\
\
Voice synthesis](https://docs.runanywhere.ai/flutter/tts/synthesize)

[Installation](https://docs.runanywhere.ai/flutter/installation)
[chat()](https://docs.runanywhere.ai/flutter/llm/chat)

⌘I

---

# Introduction - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/introduction#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Kotlin SDK

Introduction

On this page

*   [Overview](https://docs.runanywhere.ai/kotlin/introduction#overview)
    
*   [Key Capabilities](https://docs.runanywhere.ai/kotlin/introduction#key-capabilities)
    
*   [Core Philosophy](https://docs.runanywhere.ai/kotlin/introduction#core-philosophy)
    
*   [Features](https://docs.runanywhere.ai/kotlin/introduction#features)
    
*   [Language Models (LLM)](https://docs.runanywhere.ai/kotlin/introduction#language-models-llm)
    
*   [Speech-to-Text (STT)](https://docs.runanywhere.ai/kotlin/introduction#speech-to-text-stt)
    
*   [Text-to-Speech (TTS)](https://docs.runanywhere.ai/kotlin/introduction#text-to-speech-tts)
    
*   [Voice Activity Detection (VAD)](https://docs.runanywhere.ai/kotlin/introduction#voice-activity-detection-vad)
    
*   [Voice Agent Pipeline](https://docs.runanywhere.ai/kotlin/introduction#voice-agent-pipeline)
    
*   [System Requirements](https://docs.runanywhere.ai/kotlin/introduction#system-requirements)
    
*   [SDK Modules](https://docs.runanywhere.ai/kotlin/introduction#sdk-modules)
    
*   [Architecture](https://docs.runanywhere.ai/kotlin/introduction#architecture)
    
*   [Next Steps](https://docs.runanywhere.ai/kotlin/introduction#next-steps)
    

[​](https://docs.runanywhere.ai/kotlin/introduction#overview)

Overview
-------------------------------------------------------------------------

The **RunAnywhere Kotlin SDK** is a production-grade, on-device AI SDK for Android. It enables developers to run AI models directly on Android devices without requiring network connectivity for inference, ensuring minimal latency and maximum privacy for your users. The SDK provides a unified interface to multiple AI capabilities:

[LLM\
---\
\
Text generation with streaming support via Kotlin Flows](https://docs.runanywhere.ai/kotlin/llm/chat)
[STT\
---\
\
Speech-to-text transcription with Whisper models](https://docs.runanywhere.ai/kotlin/stt/transcribe)
[TTS\
---\
\
Neural voice synthesis with Sherpa-ONNX](https://docs.runanywhere.ai/kotlin/tts/synthesize)
[VAD\
---\
\
Real-time voice activity detection](https://docs.runanywhere.ai/kotlin/vad)

[​](https://docs.runanywhere.ai/kotlin/introduction#key-capabilities)

Key Capabilities
-----------------------------------------------------------------------------------------

*   **Multi-backend architecture** – Choose from LlamaCPP (GGUF models) or ONNX Runtime
*   **GPU acceleration** – Hardware-accelerated inference on supported devices
*   **Kotlin-first** – Built with Coroutines and Flows for async operations
*   **Production-ready** – Built-in analytics, logging, and model lifecycle management

[​](https://docs.runanywhere.ai/kotlin/introduction#core-philosophy)

Core Philosophy
---------------------------------------------------------------------------------------

On-Device First

All AI inference runs locally, ensuring low latency and data privacy. Once models are downloaded, no network connection is required for inference.

Modular Architecture

Backend engines are optional modules—include only what you need. This keeps your APK size minimal.

Privacy by Design

Audio and text data never leaves the device unless explicitly configured. Only anonymous analytics are collected by default.

Platform Parity

API mirrors the iOS Swift SDK for consistent cross-platform development.

[​](https://docs.runanywhere.ai/kotlin/introduction#features)

Features
-------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/kotlin/introduction#language-models-llm)

Language Models (LLM)

*   On-device text generation with streaming support
*   Kotlin Flow-based token streaming
*   System prompts and customizable generation parameters
*   Support for thinking/reasoning models
*   LlamaCPP backend for GGUF models

### 

[​](https://docs.runanywhere.ai/kotlin/introduction#speech-to-text-stt)

Speech-to-Text (STT)

*   Real-time streaming transcription
*   Batch audio transcription
*   Multi-language support
*   Whisper-based models via ONNX Runtime

### 

[​](https://docs.runanywhere.ai/kotlin/introduction#text-to-speech-tts)

Text-to-Speech (TTS)

*   Neural voice synthesis with Sherpa-ONNX
*   System voices via Android TTS
*   Streaming audio generation for long text
*   Customizable voice, pitch, rate, and volume

### 

[​](https://docs.runanywhere.ai/kotlin/introduction#voice-activity-detection-vad)

Voice Activity Detection (VAD)

*   Energy-based speech detection with Silero VAD
*   Configurable sensitivity thresholds
*   Real-time audio stream processing

### 

[​](https://docs.runanywhere.ai/kotlin/introduction#voice-agent-pipeline)

Voice Agent Pipeline

*   Full VAD → STT → LLM → TTS orchestration
*   Complete voice conversation flow
*   Streaming and batch processing modes

[​](https://docs.runanywhere.ai/kotlin/introduction#system-requirements)

System Requirements
-----------------------------------------------------------------------------------------------

| Platform | Minimum Version |
| --- | --- |
| Android | API 24 (7.0+) |

**Kotlin:** 2.0+ **Gradle:** 8.0+

ARM64 devices are recommended for best performance. GPU acceleration provides significant speedups on supported devices.

[​](https://docs.runanywhere.ai/kotlin/introduction#sdk-modules)

SDK Modules
-------------------------------------------------------------------------------

| Module | Purpose |
| --- | --- |
| `runanywhere-core` | Core SDK (required) |
| `runanywhere-llamacpp` | LLM text generation with GGUF models |
| `runanywhere-onnx` | STT/TTS/VAD via ONNX Runtime |

[​](https://docs.runanywhere.ai/kotlin/introduction#architecture)

Architecture
---------------------------------------------------------------------------------

[​](https://docs.runanywhere.ai/kotlin/introduction#next-steps)

Next Steps
-----------------------------------------------------------------------------

[Installation\
------------\
\
Add the SDK to your project via Gradle](https://docs.runanywhere.ai/kotlin/installation)
[Quick Start\
-----------\
\
Build your first AI feature in minutes](https://docs.runanywhere.ai/kotlin/quick-start)

[Best Practices](https://docs.runanywhere.ai/swift/best-practices)
[Installation](https://docs.runanywhere.ai/kotlin/installation)

⌘I

---

# transcribe() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/stt/transcribe#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

transcribe()

On this page

*   [Basic Usage](https://docs.runanywhere.ai/flutter/stt/transcribe#basic-usage)
    
*   [With Detailed Result](https://docs.runanywhere.ai/flutter/stt/transcribe#with-detailed-result)
    
*   [STTResult](https://docs.runanywhere.ai/flutter/stt/transcribe#sttresult)
    
*   [Setup](https://docs.runanywhere.ai/flutter/stt/transcribe#setup)
    
*   [1\. Register ONNX Backend](https://docs.runanywhere.ai/flutter/stt/transcribe#1-register-onnx-backend)
    
*   [2\. Add STT Model](https://docs.runanywhere.ai/flutter/stt/transcribe#2-add-stt-model)
    
*   [3\. Download & Load](https://docs.runanywhere.ai/flutter/stt/transcribe#3-download-%26-load)
    
*   [Audio Format Requirements](https://docs.runanywhere.ai/flutter/stt/transcribe#audio-format-requirements)
    
*   [Recording Audio](https://docs.runanywhere.ai/flutter/stt/transcribe#recording-audio)
    
*   [Available Models](https://docs.runanywhere.ai/flutter/stt/transcribe#available-models)
    
*   [Complete Example](https://docs.runanywhere.ai/flutter/stt/transcribe#complete-example)
    
*   [See Also](https://docs.runanywhere.ai/flutter/stt/transcribe#see-also)
    

Transcribe audio data to text using on-device Whisper models via ONNX Runtime.

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#basic-usage)

Basic Usage
----------------------------------------------------------------------------------

Copy

Ask AI

    // Load STT model first
    await RunAnywhere.loadSTTModel('sherpa-onnx-whisper-tiny.en');
    
    // Transcribe audio data (PCM16 at 16kHz mono)
    final text = await RunAnywhere.transcribe(audioBytes);
    print('Transcription: $text');
    

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#with-detailed-result)

With Detailed Result
----------------------------------------------------------------------------------------------------

Get confidence scores and metadata with `transcribeWithResult()`:

Copy

Ask AI

    final result = await RunAnywhere.transcribeWithResult(audioBytes);
    
    print('Text: ${result.text}');
    print('Confidence: ${(result.confidence * 100).toStringAsFixed(1)}%');
    print('Duration: ${result.durationMs}ms');
    if (result.language != null) {
      print('Language: ${result.language}');
    }
    

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#sttresult)

STTResult
------------------------------------------------------------------------------

| Property | Type | Description |
| --- | --- | --- |
| `text` | `String` | Transcribed text |
| `confidence` | `double` | Confidence score (0.0 to 1.0) |
| `durationMs` | `int` | Audio duration in milliseconds |
| `language` | `String?` | Detected language code |

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#setup)

Setup
----------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#1-register-onnx-backend)

1\. Register ONNX Backend

Copy

Ask AI

    import 'package:runanywhere_onnx/runanywhere_onnx.dart';
    
    await Onnx.register();
    

### 

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#2-add-stt-model)

2\. Add STT Model

Copy

Ask AI

    Onnx.addModel(
      id: 'sherpa-onnx-whisper-tiny.en',
      name: 'Whisper Tiny English',
      url: 'https://github.com/RunanywhereAI/sherpa-onnx/releases/download/runanywhere-models-v1/sherpa-onnx-whisper-tiny.en.tar.gz',
      modality: ModelCategory.speechRecognition,
    );
    

### 

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#3-download-&-load)

3\. Download & Load

Copy

Ask AI

    // Download
    await for (final progress in RunAnywhere.downloadModel('sherpa-onnx-whisper-tiny.en')) {
      print('${(progress.percentage * 100).toStringAsFixed(1)}%');
      if (progress.state.isCompleted) break;
    }
    
    // Load
    await RunAnywhere.loadSTTModel('sherpa-onnx-whisper-tiny.en');
    

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#audio-format-requirements)

Audio Format Requirements
--------------------------------------------------------------------------------------------------------------

Audio must be in the correct format for accurate transcription.

| Property | Required Value |
| --- | --- |
| Format | PCM (raw audio) |
| Sample Rate | 16,000 Hz (16kHz) |
| Channels | 1 (mono) |
| Bit Depth | 16-bit |

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#recording-audio)

Recording Audio
------------------------------------------------------------------------------------------

Use a package like `record` to capture audio:

Copy

Ask AI

    import 'package:record/record.dart';
    
    final recorder = AudioRecorder();
    
    // Start recording with correct format
    await recorder.start(
      const RecordConfig(
        encoder: AudioEncoder.pcm16bits,
        sampleRate: 16000,
        numChannels: 1,
      ),
      path: tempFilePath,
    );
    
    // Stop and get audio
    await recorder.stop();
    final audioBytes = await File(tempFilePath).readAsBytes();
    
    // Transcribe
    final text = await RunAnywhere.transcribe(audioBytes);
    

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#available-models)

Available Models
--------------------------------------------------------------------------------------------

| Model ID | Size | Languages | Speed |
| --- | --- | --- | --- |
| `sherpa-onnx-whisper-tiny.en` | ~75MB | English | Fast |
| `sherpa-onnx-whisper-base.en` | ~150MB | English | Medium |
| `sherpa-onnx-whisper-small.en` | ~250MB | English | Slower |
| `sherpa-onnx-whisper-tiny` | ~75MB | Multilingual | Fast |

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#complete-example)

Complete Example
--------------------------------------------------------------------------------------------

Copy

Ask AI

    class TranscriptionDemo extends StatefulWidget {
      @override
      _TranscriptionDemoState createState() => _TranscriptionDemoState();
    }
    
    class _TranscriptionDemoState extends State<TranscriptionDemo> {
      final _recorder = AudioRecorder();
      String _transcription = '';
      bool _isRecording = false;
      bool _isTranscribing = false;
    
      Future<void> _toggleRecording() async {
        if (_isRecording) {
          // Stop and transcribe
          final path = await _recorder.stop();
          if (path != null) {
            setState(() {
              _isRecording = false;
              _isTranscribing = true;
            });
    
            final audioBytes = await File(path).readAsBytes();
            final text = await RunAnywhere.transcribe(audioBytes);
    
            setState(() {
              _transcription = text;
              _isTranscribing = false;
            });
          }
        } else {
          // Start recording
          if (await _recorder.hasPermission()) {
            final tempDir = await getTemporaryDirectory();
            final path = '${tempDir.path}/recording.pcm';
    
            await _recorder.start(
              const RecordConfig(
                encoder: AudioEncoder.pcm16bits,
                sampleRate: 16000,
                numChannels: 1,
              ),
              path: path,
            );
    
            setState(() => _isRecording = true);
          }
        }
      }
    
      @override
      Widget build(BuildContext context) {
        return Column(
          children: [\
            Text(_transcription),\
            ElevatedButton(\
              onPressed: _isTranscribing ? null : _toggleRecording,\
              child: Text(_isRecording ? 'Stop' : 'Record'),\
            ),\
            if (_isTranscribing) CircularProgressIndicator(),\
          ],
        );
      }
    }
    

[​](https://docs.runanywhere.ai/flutter/stt/transcribe#see-also)

See Also
----------------------------------------------------------------------------

[Streaming STT\
-------------\
\
Real-time transcription](https://docs.runanywhere.ai/flutter/stt/stream)
[STT Options\
-----------\
\
Configuration options](https://docs.runanywhere.ai/flutter/stt/options)

[System Prompts](https://docs.runanywhere.ai/flutter/llm/system-prompts)
[Streaming STT](https://docs.runanywhere.ai/flutter/stt/stream)

⌘I

---

# Error Handling - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/error-handling#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Kotlin SDK

Error Handling

On this page

*   [SDKError](https://docs.runanywhere.ai/kotlin/error-handling#sdkerror)
    
*   [Basic Error Handling](https://docs.runanywhere.ai/kotlin/error-handling#basic-error-handling)
    
*   [Error Categories](https://docs.runanywhere.ai/kotlin/error-handling#error-categories)
    
*   [Error Categories Reference](https://docs.runanywhere.ai/kotlin/error-handling#error-categories-reference)
    
*   [Common Error Codes](https://docs.runanywhere.ai/kotlin/error-handling#common-error-codes)
    
*   [Error Factory Methods](https://docs.runanywhere.ai/kotlin/error-handling#error-factory-methods)
    
*   [Example: Comprehensive Error Handler](https://docs.runanywhere.ai/kotlin/error-handling#example-comprehensive-error-handler)
    
*   [Subscribe to Error Events](https://docs.runanywhere.ai/kotlin/error-handling#subscribe-to-error-events)
    

[​](https://docs.runanywhere.ai/kotlin/error-handling#sdkerror)

SDKError
---------------------------------------------------------------------------

All SDK operations throw `SDKError` for error conditions:

Copy

Ask AI

    data class SDKError(
        val code: ErrorCode,           // Specific error type
        val category: ErrorCategory,   // Error group
        val message: String,           // Human-readable message
        val cause: Throwable?          // Underlying exception
    ) : Exception(message, cause)
    

[​](https://docs.runanywhere.ai/kotlin/error-handling#basic-error-handling)

Basic Error Handling
---------------------------------------------------------------------------------------------------

Copy

Ask AI

    try {
        RunAnywhere.loadLLMModel("my-model")
        val result = RunAnywhere.generate("Hello")
    
    } catch (e: SDKError) {
        when (e.code) {
            ErrorCode.NOT_INITIALIZED -> {
                showError("Please restart the app")
            }
            ErrorCode.MODEL_NOT_FOUND -> {
                showError("Model not found. Please download it first.")
            }
            ErrorCode.MODEL_LOAD_FAILED -> {
                showError("Failed to load model: ${e.message}")
            }
            ErrorCode.LLM_GENERATION_FAILED -> {
                showError("Generation failed: ${e.message}")
            }
            ErrorCode.INSUFFICIENT_STORAGE -> {
                showError("Not enough storage space")
            }
            else -> {
                showError(e.message ?: "Unknown error")
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/error-handling#error-categories)

Error Categories
-------------------------------------------------------------------------------------------

Handle errors by category for broader matching:

Copy

Ask AI

    try {
        // SDK operations...
    } catch (e: SDKError) {
        when (e.category) {
            ErrorCategory.INITIALIZATION -> {
                // SDK not properly initialized
                restartApp()
            }
            ErrorCategory.MODEL -> {
                // Model-related issues
                promptModelDownload()
            }
            ErrorCategory.LLM -> {
                // Text generation errors
                retryWithSmallerPrompt()
            }
            ErrorCategory.STT -> {
                // Speech recognition errors
                checkMicrophonePermission()
            }
            ErrorCategory.TTS -> {
                // Speech synthesis errors
                checkAudioOutput()
            }
            ErrorCategory.NETWORK -> {
                // Network connectivity issues
                showOfflineMessage()
            }
            ErrorCategory.STORAGE -> {
                // Storage/file system errors
                promptStorageCleanup()
            }
            else -> {
                logError(e)
                showGenericError()
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/error-handling#error-categories-reference)

Error Categories Reference
---------------------------------------------------------------------------------------------------------------

| Category | Description | Common Errors |
| --- | --- | --- |
| `INITIALIZATION` | SDK startup | `NOT_INITIALIZED`, `ALREADY_INITIALIZED` |
| `MODEL` | Model operations | `MODEL_NOT_FOUND`, `MODEL_LOAD_FAILED` |
| `LLM` | Text generation | `LLM_GENERATION_FAILED` |
| `STT` | Speech-to-text | `STT_TRANSCRIPTION_FAILED` |
| `TTS` | Text-to-speech | `TTS_SYNTHESIS_FAILED` |
| `VAD` | Voice detection | `VAD_DETECTION_FAILED` |
| `NETWORK` | Network issues | `NETWORK_UNAVAILABLE`, `TIMEOUT` |
| `DOWNLOAD` | Model downloads | `DOWNLOAD_FAILED`, `DOWNLOAD_CANCELLED` |
| `STORAGE` | Storage issues | `INSUFFICIENT_STORAGE`, `FILE_NOT_FOUND` |

[​](https://docs.runanywhere.ai/kotlin/error-handling#common-error-codes)

Common Error Codes
-----------------------------------------------------------------------------------------------

Copy

Ask AI

    enum class ErrorCode {
        // General
        SUCCESS,
        UNKNOWN,
        INVALID_ARGUMENT,
    
        // Initialization
        NOT_INITIALIZED,
        ALREADY_INITIALIZED,
    
        // Model
        MODEL_NOT_FOUND,
        MODEL_NOT_LOADED,
        MODEL_LOAD_FAILED,
    
        // LLM
        LLM_GENERATION_FAILED,
    
        // STT
        STT_TRANSCRIPTION_FAILED,
    
        // TTS
        TTS_SYNTHESIS_FAILED,
    
        // VAD
        VAD_DETECTION_FAILED,
    
        // Network
        NETWORK_ERROR,
        NETWORK_UNAVAILABLE,
        TIMEOUT,
    
        // Download
        DOWNLOAD_FAILED,
        DOWNLOAD_CANCELLED,
    
        // Storage
        INSUFFICIENT_STORAGE,
        FILE_NOT_FOUND,
        OUT_OF_MEMORY
    }
    

[​](https://docs.runanywhere.ai/kotlin/error-handling#error-factory-methods)

Error Factory Methods
-----------------------------------------------------------------------------------------------------

Create specific errors programmatically:

Copy

Ask AI

    // These are used internally by the SDK
    SDKError.modelNotFound(modelId)
    SDKError.modelNotLoaded(modelId)
    SDKError.modelLoadFailed(modelId, reason)
    SDKError.llmGenerationFailed(reason)
    SDKError.sttTranscriptionFailed(reason)
    SDKError.ttsSynthesisFailed(reason)
    SDKError.networkUnavailable()
    SDKError.timeout(operation, timeoutMs)
    SDKError.insufficientStorage(requiredBytes)
    SDKError.fileNotFound(path)
    

[​](https://docs.runanywhere.ai/kotlin/error-handling#example-comprehensive-error-handler)

Example: Comprehensive Error Handler
----------------------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class SDKErrorHandler {
    
        fun handle(error: SDKError, context: Context) {
            Log.e("SDK", "Error: ${error.code} - ${error.message}", error.cause)
    
            val userMessage = when (error.code) {
                // Initialization errors
                ErrorCode.NOT_INITIALIZED ->
                    "The AI system isn't ready yet. Please wait or restart the app."
    
                // Model errors
                ErrorCode.MODEL_NOT_FOUND ->
                    "This model isn't available. Please download it first."
                ErrorCode.MODEL_LOAD_FAILED ->
                    "Couldn't load the AI model. Try freeing up some memory."
    
                // Generation errors
                ErrorCode.LLM_GENERATION_FAILED ->
                    "Couldn't generate a response. Please try again."
                ErrorCode.STT_TRANSCRIPTION_FAILED ->
                    "Couldn't understand the audio. Please speak clearly."
                ErrorCode.TTS_SYNTHESIS_FAILED ->
                    "Couldn't create audio. Please try again."
    
                // Network errors
                ErrorCode.NETWORK_UNAVAILABLE ->
                    "No internet connection. Some features may be limited."
                ErrorCode.TIMEOUT ->
                    "Request took too long. Please try again."
    
                // Download errors
                ErrorCode.DOWNLOAD_FAILED ->
                    "Download failed. Please check your connection and try again."
                ErrorCode.DOWNLOAD_CANCELLED ->
                    "Download was cancelled."
    
                // Storage errors
                ErrorCode.INSUFFICIENT_STORAGE ->
                    "Not enough storage space. Free up some space and try again."
                ErrorCode.OUT_OF_MEMORY ->
                    "Not enough memory. Try closing other apps."
    
                else ->
                    "Something went wrong. Please try again."
            }
    
            Toast.makeText(context, userMessage, Toast.LENGTH_LONG).show()
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/error-handling#subscribe-to-error-events)

Subscribe to Error Events
-------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    lifecycleScope.launch {
        RunAnywhere.events.errorEvents.collect { errorEvent ->
            Log.e("SDK", "Error event: ${errorEvent.code} - ${errorEvent.message}")
            // Handle or report error
        }
    }
    

[Configuration](https://docs.runanywhere.ai/kotlin/configuration)
[Best Practices](https://docs.runanywhere.ai/kotlin/best-practices)

⌘I

---

# chat() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/llm/chat#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

chat()

On this page

*   [When to Use](https://docs.runanywhere.ai/kotlin/llm/chat#when-to-use)
    
*   [Example: Simple Q&A](https://docs.runanywhere.ai/kotlin/llm/chat#example-simple-q%26a)
    
*   [Example: Chat Bot](https://docs.runanywhere.ai/kotlin/llm/chat#example-chat-bot)
    

The simplest way to generate text—a single function call that returns the response directly.

Copy

Ask AI

    // One-liner for quick responses
    val response = RunAnywhere.chat("What is the capital of France?")
    println(response)  // "The capital of France is Paris."
    

[​](https://docs.runanywhere.ai/kotlin/llm/chat#when-to-use)

When to Use
---------------------------------------------------------------------------

Use `chat()` when you need:

*   Quick, simple responses
*   No metrics or metadata
*   Minimal code

For more control, use [`generate()`](https://docs.runanywhere.ai/kotlin/llm/generate)
 or [`generateStream()`](https://docs.runanywhere.ai/kotlin/llm/stream)
.

[​](https://docs.runanywhere.ai/kotlin/llm/chat#example-simple-q&a)

Example: Simple Q&A
------------------------------------------------------------------------------------------

Copy

Ask AI

    lifecycleScope.launch {
        try {
            val answer = RunAnywhere.chat("Explain machine learning in one sentence")
            textView.text = answer
        } catch (e: SDKError) {
            showError(e.message)
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/llm/chat#example-chat-bot)

Example: Chat Bot
--------------------------------------------------------------------------------------

Copy

Ask AI

    suspend fun sendMessage(userMessage: String): String {
        // Ensure model is loaded
        if (!RunAnywhere.isLLMModelLoaded()) {
            RunAnywhere.loadLLMModel(modelId)
        }
    
        return RunAnywhere.chat(userMessage)
    }
    

`chat()` uses default generation options. For custom temperature, max tokens, or system prompts, use [`generate()`](https://docs.runanywhere.ai/kotlin/llm/generate)
 instead.

[Quick Start](https://docs.runanywhere.ai/kotlin/quick-start)
[generate()](https://docs.runanywhere.ai/kotlin/llm/generate)

⌘I

---

# generate() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/llm/generate#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

generate()

On this page

*   [LLMGenerationResult](https://docs.runanywhere.ai/kotlin/llm/generate#llmgenerationresult)
    
*   [LLMGenerationOptions](https://docs.runanywhere.ai/kotlin/llm/generate#llmgenerationoptions)
    
*   [Example: Creative Writing](https://docs.runanywhere.ai/kotlin/llm/generate#example-creative-writing)
    
*   [Example: Factual Response](https://docs.runanywhere.ai/kotlin/llm/generate#example-factual-response)
    
*   [Cancel Generation](https://docs.runanywhere.ai/kotlin/llm/generate#cancel-generation)
    

Use `generate()` for full control over text generation with detailed performance metrics.

Copy

Ask AI

    val result = RunAnywhere.generate(
        prompt = "Write a haiku about Kotlin programming",
        options = LLMGenerationOptions(
            maxTokens = 50,
            temperature = 1.0f,
            topP = 0.9f,
            stopSequences = listOf("###")
        )
    )
    
    println("Response: ${result.text}")
    println("Model: ${result.modelUsed}")
    println("Tokens: ${result.tokensUsed}")
    println("Speed: ${result.tokensPerSecond} tok/s")
    println("Latency: ${result.latencyMs}ms")
    
    // For reasoning models (e.g., models with thinking capability)
    result.thinkingContent?.let { thinking ->
        println("Reasoning: $thinking")
    }
    

[​](https://docs.runanywhere.ai/kotlin/llm/generate#llmgenerationresult)

LLMGenerationResult
-----------------------------------------------------------------------------------------------

The result object contains comprehensive metrics:

| Property | Type | Description |
| --- | --- | --- |
| `text` | `String` | Generated response text |
| `thinkingContent` | `String?` | Reasoning content (for thinking models) |
| `inputTokens` | `Int` | Number of prompt tokens |
| `tokensUsed` | `Int` | Number of output tokens |
| `modelUsed` | `String` | Model ID used for generation |
| `latencyMs` | `Double` | Total generation time in milliseconds |
| `tokensPerSecond` | `Double` | Generation speed |
| `timeToFirstTokenMs` | `Double?` | Time to first token (streaming) |
| `framework` | `String?` | Inference framework used |

[​](https://docs.runanywhere.ai/kotlin/llm/generate#llmgenerationoptions)

LLMGenerationOptions
-------------------------------------------------------------------------------------------------

Customize generation behavior:

Copy

Ask AI

    data class LLMGenerationOptions(
        val maxTokens: Int = 100,           // Maximum tokens to generate
        val temperature: Float = 0.8f,      // Creativity (0.0-2.0)
        val topP: Float = 1.0f,             // Nucleus sampling (0.0-1.0)
        val stopSequences: List<String> = emptyList(),
        val streamingEnabled: Boolean = false,
        val systemPrompt: String? = null    // System behavior prompt
    )
    

[​](https://docs.runanywhere.ai/kotlin/llm/generate#example-creative-writing)

Example: Creative Writing
----------------------------------------------------------------------------------------------------------

Copy

Ask AI

    val story = RunAnywhere.generate(
        prompt = "Write a short story about a robot learning to paint",
        options = LLMGenerationOptions(
            maxTokens = 500,
            temperature = 1.2f,   // Higher = more creative
            topP = 0.95f
        )
    )
    

[​](https://docs.runanywhere.ai/kotlin/llm/generate#example-factual-response)

Example: Factual Response
----------------------------------------------------------------------------------------------------------

Copy

Ask AI

    val facts = RunAnywhere.generate(
        prompt = "List the planets in our solar system",
        options = LLMGenerationOptions(
            maxTokens = 200,
            temperature = 0.1f,   // Lower = more deterministic
            topP = 0.5f
        )
    )
    

[​](https://docs.runanywhere.ai/kotlin/llm/generate#cancel-generation)

Cancel Generation
-------------------------------------------------------------------------------------------

Cancel an ongoing generation:

Copy

Ask AI

    // Start generation in a coroutine
    val job = lifecycleScope.launch {
        val result = RunAnywhere.generate(longPrompt, options)
    }
    
    // Cancel if needed
    cancelButton.setOnClickListener {
        RunAnywhere.cancelGeneration()
        job.cancel()
    }
    

[chat()](https://docs.runanywhere.ai/kotlin/llm/chat)
[generateStream()](https://docs.runanywhere.ai/kotlin/llm/stream)

⌘I

---

# Introduction - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/introduction#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

React Native SDK

Introduction

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/introduction#overview)
    
*   [Key Capabilities](https://docs.runanywhere.ai/react-native/introduction#key-capabilities)
    
*   [Core Philosophy](https://docs.runanywhere.ai/react-native/introduction#core-philosophy)
    
*   [Features](https://docs.runanywhere.ai/react-native/introduction#features)
    
*   [Language Models (LLM)](https://docs.runanywhere.ai/react-native/introduction#language-models-llm)
    
*   [Speech-to-Text (STT)](https://docs.runanywhere.ai/react-native/introduction#speech-to-text-stt)
    
*   [Text-to-Speech (TTS)](https://docs.runanywhere.ai/react-native/introduction#text-to-speech-tts)
    
*   [Voice Activity Detection (VAD)](https://docs.runanywhere.ai/react-native/introduction#voice-activity-detection-vad)
    
*   [Voice Agent Pipeline](https://docs.runanywhere.ai/react-native/introduction#voice-agent-pipeline)
    
*   [System Requirements](https://docs.runanywhere.ai/react-native/introduction#system-requirements)
    
*   [Multi-Package Architecture](https://docs.runanywhere.ai/react-native/introduction#multi-package-architecture)
    
*   [Architecture](https://docs.runanywhere.ai/react-native/introduction#architecture)
    
*   [Example Apps](https://docs.runanywhere.ai/react-native/introduction#example-apps)
    
*   [Next Steps](https://docs.runanywhere.ai/react-native/introduction#next-steps)
    

[​](https://docs.runanywhere.ai/react-native/introduction#overview)

Overview
-------------------------------------------------------------------------------

The **RunAnywhere React Native SDK** is a production-grade, on-device AI SDK for React Native applications. It enables developers to run AI models directly on iOS and Android devices without requiring network connectivity for inference, ensuring minimal latency and maximum privacy for your users. The SDK provides a unified interface to multiple AI capabilities:

[LLM\
---\
\
Text generation with streaming support via async iterators](https://docs.runanywhere.ai/react-native/llm/chat)
[STT\
---\
\
Speech-to-text transcription with Whisper models](https://docs.runanywhere.ai/react-native/stt/transcribe)
[TTS\
---\
\
Neural voice synthesis with Piper TTS](https://docs.runanywhere.ai/react-native/tts/synthesize)
[VAD\
---\
\
Real-time voice activity detection with Silero VAD](https://docs.runanywhere.ai/react-native/vad)

[​](https://docs.runanywhere.ai/react-native/introduction#key-capabilities)

Key Capabilities
-----------------------------------------------------------------------------------------------

*   **Multi-backend architecture** – Choose from LlamaCPP (GGUF models) or ONNX Runtime
*   **Cross-platform** – Single codebase for iOS and Android
*   **TypeScript-first** – Full type safety and IDE autocomplete
*   **Production-ready** – Built-in analytics, logging, and model lifecycle management

[​](https://docs.runanywhere.ai/react-native/introduction#core-philosophy)

Core Philosophy
---------------------------------------------------------------------------------------------

On-Device First

All AI inference runs locally, ensuring low latency and data privacy. Once models are downloaded, no network connection is required for inference.

Modular Architecture

Backend engines are optional packages—include only what you need. This keeps your app bundle size minimal.

Privacy by Design

Audio and text data never leaves the device unless explicitly configured. Only anonymous analytics are collected by default.

Platform Parity

API mirrors the iOS Swift SDK and Android Kotlin SDK for consistent cross-platform development.

[​](https://docs.runanywhere.ai/react-native/introduction#features)

Features
-------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/introduction#language-models-llm)

Language Models (LLM)

*   On-device text generation with streaming support
*   Async iterator-based token streaming
*   System prompts and customizable generation parameters
*   Support for thinking/reasoning models
*   LlamaCPP backend for GGUF models

### 

[​](https://docs.runanywhere.ai/react-native/introduction#speech-to-text-stt)

Speech-to-Text (STT)

*   Real-time streaming transcription
*   Batch audio transcription
*   Multi-language support
*   Whisper-based models via ONNX Runtime

### 

[​](https://docs.runanywhere.ai/react-native/introduction#text-to-speech-tts)

Text-to-Speech (TTS)

*   Neural voice synthesis with Piper TTS
*   System voices via platform TTS
*   Streaming audio generation for long text
*   Customizable voice, pitch, rate, and volume

### 

[​](https://docs.runanywhere.ai/react-native/introduction#voice-activity-detection-vad)

Voice Activity Detection (VAD)

*   Energy-based speech detection with Silero VAD
*   Configurable sensitivity thresholds
*   Real-time audio stream processing

### 

[​](https://docs.runanywhere.ai/react-native/introduction#voice-agent-pipeline)

Voice Agent Pipeline

*   Full VAD → STT → LLM → TTS orchestration
*   Complete voice conversation flow
*   Push-to-talk and hands-free modes

[​](https://docs.runanywhere.ai/react-native/introduction#system-requirements)

System Requirements
-----------------------------------------------------------------------------------------------------

| Platform | Minimum Version |
| --- | --- |
| React Native | 0.74+ |
| iOS | 15.1+ |
| Android | API 24 (7.0+) |

**Node.js:** 18+ **TypeScript:** 5.0+ (recommended)

Apple Silicon devices (M1/M2/M3, A14+) and Android devices with 6GB+ RAM are recommended. Metal GPU acceleration provides 3-5x speedup on iOS.

[​](https://docs.runanywhere.ai/react-native/introduction#multi-package-architecture)

Multi-Package Architecture
-------------------------------------------------------------------------------------------------------------------

| Package | Purpose |
| --- | --- |
| `@runanywhere/core` | Core SDK (required) |
| `@runanywhere/llamacpp` | LLM text generation with GGUF models |
| `@runanywhere/onnx` | STT/TTS/VAD via ONNX Runtime |

[​](https://docs.runanywhere.ai/react-native/introduction#architecture)

Architecture
---------------------------------------------------------------------------------------

[​](https://docs.runanywhere.ai/react-native/introduction#example-apps)

Example Apps
---------------------------------------------------------------------------------------

Working example apps to help you get started:

*   [**React Native Starter App**](https://github.com/RunanywhereAI/react-native-starter-app)
     – Bare React Native CLI app with Chat, STT, TTS, and Voice Pipeline demos
*   [**Monorepo Example**](https://github.com/RunanywhereAI/runanywhere-sdks/tree/main/examples/react-native/RunAnywhereAI)
     – Full-featured reference app from the SDK repository

[​](https://docs.runanywhere.ai/react-native/introduction#next-steps)

Next Steps
-----------------------------------------------------------------------------------

[Installation\
------------\
\
Add the SDK to your project via npm](https://docs.runanywhere.ai/react-native/installation)
[Quick Start\
-----------\
\
Build your first AI feature in minutes](https://docs.runanywhere.ai/react-native/quick-start)

[Best Practices](https://docs.runanywhere.ai/kotlin/best-practices)
[Installation](https://docs.runanywhere.ai/react-native/installation)

⌘I

---

# Installation - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/installation#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Swift SDK

Installation

On this page

*   [Swift Package Manager (Xcode)](https://docs.runanywhere.ai/swift/installation#swift-package-manager-xcode)
    
*   [Package.swift](https://docs.runanywhere.ai/swift/installation#package-swift)
    
*   [Available Products](https://docs.runanywhere.ai/swift/installation#available-products)
    
*   [Binary Size Impact](https://docs.runanywhere.ai/swift/installation#binary-size-impact)
    
*   [Verify Installation](https://docs.runanywhere.ai/swift/installation#verify-installation)
    
*   [Troubleshooting](https://docs.runanywhere.ai/swift/installation#troubleshooting)
    
*   [Next Steps](https://docs.runanywhere.ai/swift/installation#next-steps)
    

[​](https://docs.runanywhere.ai/swift/installation#swift-package-manager-xcode)

Swift Package Manager (Xcode)
----------------------------------------------------------------------------------------------------------------

1

[](https://docs.runanywhere.ai/swift/installation#)

Open your project

Open your project in Xcode

2

[](https://docs.runanywhere.ai/swift/installation#)

Add package dependency

Go to **File → Add Package Dependencies…**

3

[](https://docs.runanywhere.ai/swift/installation#)

Enter repository URL

Enter the repository URL: `https://github.com/RunanywhereAI/runanywhere-sdks`

4

[](https://docs.runanywhere.ai/swift/installation#)

Select version

Select the version (e.g., `from: "0.16.0"`)

5

[](https://docs.runanywhere.ai/swift/installation#)

Choose products

Choose the products you need: - **RunAnywhere** (required) – Core SDK - **RunAnywhereLlamaCPP** – LLM text generation with GGUF models - **RunAnywhereONNX** – ONNX Runtime for STT/TTS/VAD

[​](https://docs.runanywhere.ai/swift/installation#package-swift)

Package.swift
----------------------------------------------------------------------------------

For Swift packages, add the dependency directly to your `Package.swift`:

Package.swift

Copy

Ask AI

    dependencies: [\
        .package(url: "https://github.com/RunanywhereAI/runanywhere-sdks", from: "0.16.0")\
    ]
    

Then add the products you need to your target dependencies:

Package.swift

Copy

Ask AI

    targets: [\
        .target(\
            name: "YourApp",\
            dependencies: [\
                .product(name: "RunAnywhere", package: "runanywhere-sdks"),\
                .product(name: "RunAnywhereLlamaCPP", package: "runanywhere-sdks"),\
                .product(name: "RunAnywhereONNX", package: "runanywhere-sdks"),\
            ]\
        )\
    ]
    

[​](https://docs.runanywhere.ai/swift/installation#available-products)

Available Products
--------------------------------------------------------------------------------------------

| Product | Description |
| --- | --- |
| `RunAnywhere` | Core SDK (required for all features) |
| `RunAnywhereLlamaCPP` | LLM capability with GGUF models |
| `RunAnywhereONNX` | STT, TTS, and VAD via ONNX Runtime |

Only include the products you need to minimize your app’s binary size. The modular architecture means you can add capabilities incrementally.

[​](https://docs.runanywhere.ai/swift/installation#binary-size-impact)

Binary Size Impact
--------------------------------------------------------------------------------------------

| Configuration | Approximate Size |
| --- | --- |
| LLM only | ~17 MB |
| STT only (ONNX) | ~52 MB |
| LLM + STT + TTS | ~87 MB |
| Full (all backends) | ~95 MB |

[​](https://docs.runanywhere.ai/swift/installation#verify-installation)

Verify Installation
----------------------------------------------------------------------------------------------

After installation, verify everything is working:

Copy

Ask AI

    import RunAnywhere
    
    // Check SDK version
    print("SDK Version: \(RunAnywhere.version)")
    

[​](https://docs.runanywhere.ai/swift/installation#troubleshooting)

Troubleshooting
--------------------------------------------------------------------------------------

Package resolution fails

Try resetting package caches: 1. Go to **File → Packages → Reset Package Caches** 2. Clean the build folder: **Product → Clean Build Folder** 3. Try adding the package again

Module not found errors

Ensure you’ve added the correct product to your target’s dependencies. Each module must be explicitly added.

Linker errors with XCFrameworks

If you encounter linker errors, ensure your deployment target meets the minimum requirements: - iOS 17.0+ - macOS 14.0+

[​](https://docs.runanywhere.ai/swift/installation#next-steps)

Next Steps
----------------------------------------------------------------------------

[Quick Start\
-----------\
\
Build your first AI feature →](https://docs.runanywhere.ai/swift/quick-start)

[Introduction](https://docs.runanywhere.ai/swift/introduction)
[Quick Start](https://docs.runanywhere.ai/swift/quick-start)

⌘I

---

# Best Practices - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/best-practices#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Kotlin SDK

Best Practices

On this page

*   [Memory Management](https://docs.runanywhere.ai/kotlin/best-practices#memory-management)
    
*   [Load Only What You Need](https://docs.runanywhere.ai/kotlin/best-practices#load-only-what-you-need)
    
*   [Unload When Not Needed](https://docs.runanywhere.ai/kotlin/best-practices#unload-when-not-needed)
    
*   [Monitor Memory Before Loading](https://docs.runanywhere.ai/kotlin/best-practices#monitor-memory-before-loading)
    
*   [Performance Optimization](https://docs.runanywhere.ai/kotlin/best-practices#performance-optimization)
    
*   [Use Quantized Models](https://docs.runanywhere.ai/kotlin/best-practices#use-quantized-models)
    
*   [Use Streaming for Better UX](https://docs.runanywhere.ai/kotlin/best-practices#use-streaming-for-better-ux)
    
*   [Set Appropriate Token Limits](https://docs.runanywhere.ai/kotlin/best-practices#set-appropriate-token-limits)
    
*   [App Lifecycle](https://docs.runanywhere.ai/kotlin/best-practices#app-lifecycle)
    
*   [Handle Background/Foreground](https://docs.runanywhere.ai/kotlin/best-practices#handle-background%2Fforeground)
    
*   [Preload Models at Launch](https://docs.runanywhere.ai/kotlin/best-practices#preload-models-at-launch)
    
*   [Error Handling](https://docs.runanywhere.ai/kotlin/best-practices#error-handling)
    
*   [Always Handle Errors](https://docs.runanywhere.ai/kotlin/best-practices#always-handle-errors)
    
*   [Provide User Feedback](https://docs.runanywhere.ai/kotlin/best-practices#provide-user-feedback)
    
*   [Testing](https://docs.runanywhere.ai/kotlin/best-practices#testing)
    
*   [Test on Real Devices](https://docs.runanywhere.ai/kotlin/best-practices#test-on-real-devices)
    
*   [Measure Performance](https://docs.runanywhere.ai/kotlin/best-practices#measure-performance)
    
*   [Security](https://docs.runanywhere.ai/kotlin/best-practices#security)
    
*   [Protect API Keys](https://docs.runanywhere.ai/kotlin/best-practices#protect-api-keys)
    
*   [Clear Sensitive Data](https://docs.runanywhere.ai/kotlin/best-practices#clear-sensitive-data)
    
*   [Summary Checklist](https://docs.runanywhere.ai/kotlin/best-practices#summary-checklist)
    

[​](https://docs.runanywhere.ai/kotlin/best-practices#memory-management)

Memory Management
---------------------------------------------------------------------------------------------

On-device AI models are memory-intensive. Proper memory management is critical for app stability.

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#load-only-what-you-need)

Load Only What You Need

Copy

Ask AI

    // ❌ Don't load multiple large models
    RunAnywhere.loadLLMModel("model-3b")
    RunAnywhere.loadSTTModel("whisper-large")
    
    // ✅ Load one LLM at a time, use smaller models
    RunAnywhere.loadLLMModel("model-0.5b")
    

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#unload-when-not-needed)

Unload When Not Needed

Copy

Ask AI

    // Unload models when switching tasks or backgrounding
    override fun onStop() {
        super.onStop()
        lifecycleScope.launch {
            RunAnywhere.unloadLLMModel()
            RunAnywhere.unloadSTTModel()
        }
    }
    

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#monitor-memory-before-loading)

Monitor Memory Before Loading

Copy

Ask AI

    val modelInfo = RunAnywhere.model(modelId)
    val requiredMemory = modelInfo?.downloadSize ?: 0
    
    val activityManager = getSystemService(Context.ACTIVITY_SERVICE) as ActivityManager
    val memoryInfo = ActivityManager.MemoryInfo()
    activityManager.getMemoryInfo(memoryInfo)
    
    if (memoryInfo.availMem < requiredMemory * 1.5) {
        showWarning("Low memory - performance may be affected")
    }
    

[​](https://docs.runanywhere.ai/kotlin/best-practices#performance-optimization)

Performance Optimization
-----------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#use-quantized-models)

Use Quantized Models

| Model Type | Size | Performance | Quality |
| --- | --- | --- | --- |
| Q8 (8-bit) | Larger | Slower | Best |
| Q4 (4-bit) | Smaller | Faster | Good |
| Q2 (2-bit) | Smallest | Fastest | Acceptable |

Copy

Ask AI

    // ✅ Use Q4 models for mobile devices
    val model = RunAnywhere.registerModel(
        name = "Qwen 0.5B Q4",
        url = "...qwen2.5-0.5b-instruct-q4_0.gguf",
        framework = InferenceFramework.LLAMA_CPP
    )
    

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#use-streaming-for-better-ux)

Use Streaming for Better UX

Copy

Ask AI

    // ✅ Stream tokens for perceived faster responses
    RunAnywhere.generateStream(prompt)
        .collect { token ->
            textView.append(token)
        }
    
    // ❌ Waiting for full response feels slow
    val result = RunAnywhere.generate(prompt)
    textView.text = result.text
    

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#set-appropriate-token-limits)

Set Appropriate Token Limits

Copy

Ask AI

    // ✅ Match maxTokens to your use case
    val shortAnswer = LLMGenerationOptions(maxTokens = 50)   // Quick Q&A
    val mediumAnswer = LLMGenerationOptions(maxTokens = 200) // General chat
    val longForm = LLMGenerationOptions(maxTokens = 500)     // Stories/articles
    

[​](https://docs.runanywhere.ai/kotlin/best-practices#app-lifecycle)

App Lifecycle
-------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#handle-background/foreground)

Handle Background/Foreground

Copy

Ask AI

    class MyApplication : Application(), LifecycleEventObserver {
    
        override fun onCreate() {
            super.onCreate()
            ProcessLifecycleOwner.get().lifecycle.addObserver(this)
            RunAnywhere.initialize(environment = SDKEnvironment.PRODUCTION)
        }
    
        override fun onStateChanged(source: LifecycleOwner, event: Lifecycle.Event) {
            when (event) {
                Lifecycle.Event.ON_STOP -> {
                    // App backgrounded - release resources
                    CoroutineScope(Dispatchers.Main).launch {
                        RunAnywhere.stopVoiceSession()
                        RunAnywhere.cleanup()
                    }
                }
                Lifecycle.Event.ON_START -> {
                    // App foregrounded - reinitialize if needed
                }
                else -> {}
            }
        }
    }
    

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#preload-models-at-launch)

Preload Models at Launch

Copy

Ask AI

    class SplashActivity : AppCompatActivity() {
    
        override fun onCreate(savedInstanceState: Bundle?) {
            super.onCreate(savedInstanceState)
    
            lifecycleScope.launch {
                // Show loading UI
                showLoading()
    
                // Preload commonly used models
                val models = listOf("qwen-0.5b", "whisper-tiny")
    
                models.forEach { modelId ->
                    if (!RunAnywhere.isModelDownloaded(modelId)) {
                        RunAnywhere.downloadModel(modelId).collect { progress ->
                            updateProgress(modelId, progress.progress)
                        }
                    }
                }
    
                // Pre-load the primary model
                RunAnywhere.loadLLMModel("qwen-0.5b")
    
                // Navigate to main screen
                startActivity(Intent(this@SplashActivity, MainActivity::class.java))
                finish()
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/best-practices#error-handling)

Error Handling
---------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#always-handle-errors)

Always Handle Errors

Copy

Ask AI

    // ✅ Comprehensive error handling
    lifecycleScope.launch {
        try {
            val result = RunAnywhere.generate(prompt)
            showResponse(result.text)
        } catch (e: SDKError) {
            when (e.category) {
                ErrorCategory.MODEL -> promptModelDownload()
                ErrorCategory.STORAGE -> promptStorageCleanup()
                else -> showGenericError(e.message)
            }
        }
    }
    

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#provide-user-feedback)

Provide User Feedback

Copy

Ask AI

    // ✅ Show meaningful progress
    RunAnywhere.downloadModel(modelId).collect { progress ->
        when (progress.state) {
            DownloadState.DOWNLOADING -> {
                progressBar.progress = (progress.progress * 100).toInt()
                statusText.text = "Downloading AI model..."
            }
            DownloadState.EXTRACTING -> {
                statusText.text = "Preparing model..."
            }
            DownloadState.COMPLETED -> {
                statusText.text = "Ready!"
            }
            DownloadState.ERROR -> {
                showError("Download failed: ${progress.error}")
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/best-practices#testing)

Testing
-------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#test-on-real-devices)

Test on Real Devices

Copy

Ask AI

    // Simulators don't accurately reflect:
    // - Memory constraints
    // - CPU performance
    // - Thermal throttling
    // Always test on physical devices before release
    

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#measure-performance)

Measure Performance

Copy

Ask AI

    val result = RunAnywhere.generate(prompt, options)
    
    Log.d("Performance", """
        Model: ${result.modelUsed}
        Tokens: ${result.tokensUsed}
        Speed: ${result.tokensPerSecond} tok/s
        Latency: ${result.latencyMs}ms
        TTFT: ${result.timeToFirstTokenMs}ms
    """.trimIndent())
    

[​](https://docs.runanywhere.ai/kotlin/best-practices#security)

Security
---------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#protect-api-keys)

Protect API Keys

Copy

Ask AI

    // ❌ Don't hardcode API keys
    RunAnywhere.initialize(apiKey = "sk-12345...")
    
    // ✅ Use BuildConfig or secure storage
    RunAnywhere.initialize(apiKey = BuildConfig.RUNANYWHERE_API_KEY)
    

### 

[​](https://docs.runanywhere.ai/kotlin/best-practices#clear-sensitive-data)

Clear Sensitive Data

Copy

Ask AI

    // Clear conversation history when appropriate
    RunAnywhere.clearVoiceConversation()
    
    // Reset SDK to clear all state
    RunAnywhere.reset()
    

[​](https://docs.runanywhere.ai/kotlin/best-practices#summary-checklist)

Summary Checklist
---------------------------------------------------------------------------------------------

*   [ ]  Use quantized models (Q4) for mobile devices
*   [ ]  Unload models when backgrounding the app
*   [ ]  Use streaming for long text generation
*   [ ]  Handle all error categories appropriately
*   [ ]  Test on physical devices, not simulators
*   [ ]  Preload commonly used models at app startup
*   [ ]  Monitor memory before loading large models
*   [ ]  Secure API keys using BuildConfig or secure storage

[Error Handling](https://docs.runanywhere.ai/kotlin/error-handling)
[Introduction](https://docs.runanywhere.ai/react-native/introduction)

⌘I

---

# Streaming TTS - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/tts/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

Streaming TTS

On this page

*   [Streaming Synthesis](https://docs.runanywhere.ai/kotlin/tts/stream#streaming-synthesis)
    
*   [Example: Streaming Audio Player](https://docs.runanywhere.ai/kotlin/tts/stream#example-streaming-audio-player)
    
*   [Example: Chat Response with Streaming TTS](https://docs.runanywhere.ai/kotlin/tts/stream#example-chat-response-with-streaming-tts)
    
*   [Stop Synthesis](https://docs.runanywhere.ai/kotlin/tts/stream#stop-synthesis)
    
*   [When to Use Streaming](https://docs.runanywhere.ai/kotlin/tts/stream#when-to-use-streaming)
    
*   [Performance Characteristics](https://docs.runanywhere.ai/kotlin/tts/stream#performance-characteristics)
    

Stream audio chunks as they are synthesized, enabling playback to start before the full synthesis completes.

[​](https://docs.runanywhere.ai/kotlin/tts/stream#streaming-synthesis)

Streaming Synthesis
---------------------------------------------------------------------------------------------

Copy

Ask AI

    val longText = """
        This is a very long text that will be synthesized in chunks.
        Each chunk will be delivered as soon as it's ready,
        allowing playback to start immediately without waiting
        for the entire synthesis to complete.
    """.trimIndent()
    
    RunAnywhere.synthesizeStream(
        text = longText,
        options = TTSOptions(rate = 1.0f),
        onAudioChunk = { chunk ->
            // Play each chunk as it arrives
            audioPlayer.enqueue(chunk)
        }
    )
    

[​](https://docs.runanywhere.ai/kotlin/tts/stream#example-streaming-audio-player)

Example: Streaming Audio Player
--------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class StreamingAudioPlayer {
        private val audioQueue = ConcurrentLinkedQueue<ByteArray>()
        private var isPlaying = false
    
        fun enqueue(audioChunk: ByteArray) {
            audioQueue.add(audioChunk)
            if (!isPlaying) {
                startPlayback()
            }
        }
    
        private fun startPlayback() {
            isPlaying = true
            CoroutineScope(Dispatchers.IO).launch {
                while (audioQueue.isNotEmpty() || isPlaying) {
                    val chunk = audioQueue.poll()
                    if (chunk != null) {
                        playChunk(chunk)
                    } else {
                        delay(10) // Wait for more chunks
                    }
                }
            }
        }
    
        private fun playChunk(chunk: ByteArray) {
            // Play audio chunk using AudioTrack
        }
    
        fun stop() {
            isPlaying = false
            audioQueue.clear()
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/tts/stream#example-chat-response-with-streaming-tts)

Example: Chat Response with Streaming TTS
----------------------------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class ChatViewModel : ViewModel() {
        private val audioPlayer = StreamingAudioPlayer()
    
        fun sendMessageAndSpeak(prompt: String) {
            viewModelScope.launch {
                // Generate text with streaming
                val streamResult = RunAnywhere.generateStreamWithMetrics(prompt)
    
                // Collect full response for TTS
                val fullResponse = StringBuilder()
                streamResult.stream.collect { token ->
                    fullResponse.append(token)
                    _displayText.value += token
                }
    
                // Stream synthesize the full response
                RunAnywhere.synthesizeStream(
                    text = fullResponse.toString(),
                    options = TTSOptions(),
                    onAudioChunk = { chunk ->
                        audioPlayer.enqueue(chunk)
                    }
                )
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/tts/stream#stop-synthesis)

Stop Synthesis
-----------------------------------------------------------------------------------

Cancel ongoing synthesis:

Copy

Ask AI

    // Stop synthesis mid-stream
    RunAnywhere.stopSynthesis()
    

[​](https://docs.runanywhere.ai/kotlin/tts/stream#when-to-use-streaming)

When to Use Streaming
-------------------------------------------------------------------------------------------------

| Scenario | Recommended |
| --- | --- |
| Short text (under 100 chars) | `synthesize()` |
| Long text (over 200 chars) | `synthesizeStream()` |
| Real-time chat responses | `synthesizeStream()` |
| Pre-cached audio | `synthesize()` |

Streaming synthesis reduces time-to-first-audio for long text: - **Regular synthesis**: Wait for full synthesis, then play - **Streaming synthesis**: Start playing after first chunk (~100ms)

[​](https://docs.runanywhere.ai/kotlin/tts/stream#performance-characteristics)

Performance Characteristics
-------------------------------------------------------------------------------------------------------------

| Text Length | Regular Synthesis | Streaming First Audio |
| --- | --- | --- |
| 50 chars | ~100ms | ~100ms |
| 200 chars | ~400ms | ~100ms |
| 500 chars | ~1000ms | ~100ms |
| 1000 chars | ~2000ms | ~100ms |

[synthesize()](https://docs.runanywhere.ai/kotlin/tts/synthesize)
[Voices](https://docs.runanywhere.ai/kotlin/tts/voices)

⌘I

---

# SDK Overview - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/sdks#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Getting Started

SDK Overview

On this page

*   [Available SDKs](https://docs.runanywhere.ai/sdks#available-sdks)
    
*   [Feature Comparison](https://docs.runanywhere.ai/sdks#feature-comparison)
    
*   [Choosing the Right SDK](https://docs.runanywhere.ai/sdks#choosing-the-right-sdk)
    

[​](https://docs.runanywhere.ai/sdks#available-sdks)

Available SDKs
----------------------------------------------------------------------

All RunAnywhere SDKs provide the same core AI capabilities with platform-native APIs. Choose the SDK that matches your development stack.

*   Swift (Apple)
    
*   Kotlin (Android)
    
*   React Native
    
*   Flutter
    

[Swift SDK\
---------\
\
Production-grade, on-device AI for Apple platforms](https://docs.runanywhere.ai/swift/introduction)

### 

[​](https://docs.runanywhere.ai/sdks#platform-support)

Platform Support

| Platform | Minimum Version |
| --- | --- |
| iOS | 17.0+ |
| macOS | 14.0+ |
| tvOS | 17.0+ |
| watchOS | 10.0+ |

### 

[​](https://docs.runanywhere.ai/sdks#requirements)

Requirements

*   **Swift:** 5.9+
*   **Xcode:** 15.2+

### 

[​](https://docs.runanywhere.ai/sdks#key-features)

Key Features

*   Metal acceleration on Apple Silicon
*   LlamaCPP backend for GGUF models
*   ONNX Runtime for STT/TTS/VAD
*   Apple Foundation Models support (iOS 26+)

[Installation\
------------\
\
Add via Swift Package Manager](https://docs.runanywhere.ai/swift/installation)
[Quick Start\
-----------\
\
Build your first AI feature](https://docs.runanywhere.ai/swift/quick-start)

[Kotlin SDK\
----------\
\
Production-grade, on-device AI for Android](https://docs.runanywhere.ai/kotlin/introduction)

### 

[​](https://docs.runanywhere.ai/sdks#platform-support-2)

Platform Support

| Platform | Minimum Version |
| --- | --- |
| Android | API 24+ (7.0+) |

### 

[​](https://docs.runanywhere.ai/sdks#requirements-2)

Requirements

*   **Kotlin:** 2.0+
*   **Gradle:** 8.0+

### 

[​](https://docs.runanywhere.ai/sdks#key-features-2)

Key Features

*   GPU acceleration support
*   Coroutines-based async API
*   LlamaCPP backend for GGUF models
*   ONNX Runtime for STT/TTS/VAD

[Installation\
------------\
\
Add via Gradle](https://docs.runanywhere.ai/kotlin/installation)
[Quick Start\
-----------\
\
Build your first AI feature](https://docs.runanywhere.ai/kotlin/quick-start)

[React Native SDK\
----------------\
\
Cross-platform on-device AI with TypeScript](https://docs.runanywhere.ai/react-native/introduction)

### 

[​](https://docs.runanywhere.ai/sdks#platform-support-3)

Platform Support

| Platform | Minimum Version |
| --- | --- |
| iOS | 15.1+ |
| Android | API 24+ (7.0+) |

### 

[​](https://docs.runanywhere.ai/sdks#requirements-3)

Requirements

*   **React Native:** 0.71+ (recommended 0.74+)
*   **TypeScript:** 5.0+

### 

[​](https://docs.runanywhere.ai/sdks#key-features-3)

Key Features

*   TypeScript-first with full type safety
*   Single codebase for iOS and Android
*   TurboModules for optimal performance
*   LlamaCPP and ONNX backends

[Installation\
------------\
\
Add via npm/yarn](https://docs.runanywhere.ai/react-native/installation)
[Quick Start\
-----------\
\
Build your first AI feature](https://docs.runanywhere.ai/react-native/quick-start)

[Flutter SDK\
-----------\
\
Cross-platform on-device AI with Dart](https://docs.runanywhere.ai/flutter/introduction)

### 

[​](https://docs.runanywhere.ai/sdks#platform-support-4)

Platform Support

| Platform | Minimum Version |
| --- | --- |
| iOS | 14.0+ |
| Android | API 24+ (7.0+) |

### 

[​](https://docs.runanywhere.ai/sdks#requirements-4)

Requirements

*   **Flutter:** 3.10+
*   **Dart:** 3.0+

### 

[​](https://docs.runanywhere.ai/sdks#key-features-4)

Key Features

*   Dart-native async/await API
*   Single codebase for iOS and Android
*   Platform channels for native performance
*   LlamaCPP and ONNX backends

[Installation\
------------\
\
Add via pub.dev](https://docs.runanywhere.ai/flutter/installation)
[Quick Start\
-----------\
\
Build your first AI feature](https://docs.runanywhere.ai/flutter/quick-start)

* * *

[​](https://docs.runanywhere.ai/sdks#feature-comparison)

Feature Comparison
------------------------------------------------------------------------------

All SDKs have **feature parity** and provide the same core capabilities:

| Feature | Swift | Kotlin | React Native | Flutter |
| --- | --- | --- | --- | --- |
| **LLM (Text Generation)** |     |     |     |     |
| **STT (Speech-to-Text)** |     |     |     |     |
| **TTS (Text-to-Speech)** |     |     |     |     |
| **VAD (Voice Detection)** |     |     |     |     |
| **Voice Agent Pipeline** |     |     |     |     |
| **Streaming Support** |     |     |     |     |
| **Offline Capable** |     |     |     |     |
| **GPU Acceleration** | Metal | GPU | Native | Native |

* * *

[​](https://docs.runanywhere.ai/sdks#choosing-the-right-sdk)

Choosing the Right SDK
--------------------------------------------------------------------------------------

Building a native iOS/macOS app?

Use the **Swift SDK** for the best performance and access to Apple-specific features like Metal acceleration and Apple Foundation Models.[Get Started with Swift\
----------------------](https://docs.runanywhere.ai/swift/introduction)

Building a native Android app?

Use the **Kotlin SDK** for native Android performance with Coroutines-based async APIs and GPU acceleration.[Get Started with Kotlin\
-----------------------](https://docs.runanywhere.ai/kotlin/introduction)

Building a cross-platform app with JavaScript/TypeScript?

Use the **React Native SDK** if your team is already using React Native. Get full type safety with TypeScript and a single codebase.[Get Started with React Native\
-----------------------------](https://docs.runanywhere.ai/react-native/introduction)

Building a cross-platform app with Flutter/Dart?

Use the **Flutter SDK** if your team is already using Flutter. Get native Dart APIs with async/await patterns.[Get Started with Flutter\
------------------------](https://docs.runanywhere.ai/flutter/introduction)

**Not sure which to choose?** All SDKs have feature parity. Pick the one that matches your existing tech stack for the smoothest integration.

[Quick Start](https://docs.runanywhere.ai/quickstart)
[Introduction](https://docs.runanywhere.ai/swift/introduction)

⌘I

---

# System Prompts - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/llm/system-prompts#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

System Prompts

On this page

*   [Basic Usage](https://docs.runanywhere.ai/kotlin/llm/system-prompts#basic-usage)
    
*   [Example: Code Assistant](https://docs.runanywhere.ai/kotlin/llm/system-prompts#example-code-assistant)
    
*   [Example: Customer Support Bot](https://docs.runanywhere.ai/kotlin/llm/system-prompts#example-customer-support-bot)
    
*   [Example: Creative Writer](https://docs.runanywhere.ai/kotlin/llm/system-prompts#example-creative-writer)
    
*   [Example: JSON Output](https://docs.runanywhere.ai/kotlin/llm/system-prompts#example-json-output)
    
*   [Best Practices](https://docs.runanywhere.ai/kotlin/llm/system-prompts#best-practices)
    

System prompts define how the model should behave, respond, and what role it should assume.

[​](https://docs.runanywhere.ai/kotlin/llm/system-prompts#basic-usage)

Basic Usage
-------------------------------------------------------------------------------------

Copy

Ask AI

    val options = LLMGenerationOptions(
        maxTokens = 200,
        systemPrompt = """
            You are a senior Kotlin developer.
            Answer questions with code examples.
            Use modern Kotlin conventions and coroutines.
        """.trimIndent()
    )
    
    val result = RunAnywhere.generate(
        prompt = "How do I make an HTTP request?",
        options = options
    )
    

[​](https://docs.runanywhere.ai/kotlin/llm/system-prompts#example-code-assistant)

Example: Code Assistant
------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    val codeAssistant = LLMGenerationOptions(
        maxTokens = 500,
        temperature = 0.3f,  // Lower for more precise code
        systemPrompt = """
            You are an expert Android developer assistant.
            - Always provide working Kotlin code examples
            - Use Jetpack libraries when appropriate
            - Follow Material Design guidelines for UI code
            - Include error handling in your examples
            - Keep explanations concise
        """.trimIndent()
    )
    
    val result = RunAnywhere.generate(
        prompt = "How do I implement a RecyclerView with DiffUtil?",
        options = codeAssistant
    )
    

[​](https://docs.runanywhere.ai/kotlin/llm/system-prompts#example-customer-support-bot)

Example: Customer Support Bot
------------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    val supportBot = LLMGenerationOptions(
        maxTokens = 300,
        temperature = 0.7f,
        systemPrompt = """
            You are a friendly customer support assistant for TechCorp.
    
            Guidelines:
            - Be helpful, patient, and empathetic
            - Keep responses under 3 paragraphs
            - If you don't know something, say so
            - Never make up product features
            - End with asking if there's anything else you can help with
        """.trimIndent()
    )
    

[​](https://docs.runanywhere.ai/kotlin/llm/system-prompts#example-creative-writer)

Example: Creative Writer
--------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    val creativeWriter = LLMGenerationOptions(
        maxTokens = 1000,
        temperature = 1.2f,  // Higher for creativity
        topP = 0.95f,
        systemPrompt = """
            You are a creative fiction writer with a vivid imagination.
            Write engaging stories with:
            - Rich character development
            - Descriptive settings
            - Unexpected plot twists
            - Emotional depth
        """.trimIndent()
    )
    

[​](https://docs.runanywhere.ai/kotlin/llm/system-prompts#example-json-output)

Example: JSON Output
------------------------------------------------------------------------------------------------------

Copy

Ask AI

    val jsonFormatter = LLMGenerationOptions(
        maxTokens = 500,
        temperature = 0.1f,  // Very low for consistent structure
        systemPrompt = """
            You are a data extraction assistant.
            Always respond with valid JSON only.
            Do not include any text outside the JSON object.
            Use this schema: {"name": string, "category": string, "confidence": number}
        """.trimIndent()
    )
    
    val result = RunAnywhere.generate(
        prompt = "Extract product info: 'iPhone 15 Pro Max 256GB Space Black'",
        options = jsonFormatter
    )
    // Result: {"name": "iPhone 15 Pro Max", "category": "smartphone", "confidence": 0.95}
    

[​](https://docs.runanywhere.ai/kotlin/llm/system-prompts#best-practices)

Best Practices
-------------------------------------------------------------------------------------------

**Keep system prompts focused and specific:** - Define the role clearly - Set boundaries on what the model should/shouldn’t do - Specify output format if needed - Include examples for complex tasks

| Goal | Temperature | System Prompt Style |
| --- | --- | --- |
| Code generation | 0.1-0.4 | Precise, technical instructions |
| Creative writing | 0.9-1.3 | Open-ended, encourage creativity |
| Factual Q&A | 0.2-0.5 | Focus on accuracy, cite limitations |
| Conversation | 0.6-0.8 | Personality traits, response guidelines |

[generateStream()](https://docs.runanywhere.ai/kotlin/llm/stream)
[transcribe()](https://docs.runanywhere.ai/kotlin/stt/transcribe)

⌘I

---

# Streaming STT - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/stt/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

Streaming STT

On this page

*   [Streaming Transcription](https://docs.runanywhere.ai/kotlin/stt/stream#streaming-transcription)
    
*   [Real-Time Microphone Transcription](https://docs.runanywhere.ai/kotlin/stt/stream#real-time-microphone-transcription)
    
*   [Stream Audio Samples](https://docs.runanywhere.ai/kotlin/stt/stream#stream-audio-samples)
    
*   [Subscribe to STT Events](https://docs.runanywhere.ai/kotlin/stt/stream#subscribe-to-stt-events)
    
*   [Example: Voice Search](https://docs.runanywhere.ai/kotlin/stt/stream#example-voice-search)
    
*   [Performance Tips](https://docs.runanywhere.ai/kotlin/stt/stream#performance-tips)
    

Stream transcription results as audio is being processed, providing real-time feedback for voice interfaces.

[​](https://docs.runanywhere.ai/kotlin/stt/stream#streaming-transcription)

Streaming Transcription
-----------------------------------------------------------------------------------------------------

Copy

Ask AI

    val audioData = // ... recorded audio bytes
    
    RunAnywhere.transcribeStream(
        audioData = audioData,
        options = STTOptions(language = "en"),
        onPartialResult = { partial ->
            // Update UI with partial transcription
            transcriptionTextView.text = partial.transcript
        }
    )
    

[​](https://docs.runanywhere.ai/kotlin/stt/stream#real-time-microphone-transcription)

Real-Time Microphone Transcription
---------------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class LiveTranscriptionViewModel : ViewModel() {
        private val _partialText = MutableStateFlow("")
        val partialText: StateFlow<String> = _partialText
    
        private val _finalText = MutableStateFlow("")
        val finalText: StateFlow<String> = _finalText
    
        fun processAudioSamples(samples: FloatArray) {
            viewModelScope.launch {
                // Process streaming audio samples
                RunAnywhere.processStreamingAudio(samples)
            }
        }
    
        fun stopTranscription() {
            viewModelScope.launch {
                RunAnywhere.stopStreamingTranscription()
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/stt/stream#stream-audio-samples)

Stream Audio Samples
-----------------------------------------------------------------------------------------------

For continuous microphone input:

Copy

Ask AI

    // Start streaming transcription
    suspend fun startLiveTranscription(
        audioSamplesFlow: Flow<FloatArray>,
        onPartialResult: (String) -> Unit
    ) {
        audioSamplesFlow.collect { samples ->
            RunAnywhere.processStreamingAudio(samples)
            // Handle partial results via event system
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/stt/stream#subscribe-to-stt-events)

Subscribe to STT Events
-----------------------------------------------------------------------------------------------------

Use the event system for streaming updates:

Copy

Ask AI

    lifecycleScope.launch {
        RunAnywhere.events.sttEvents.collect { event ->
            when (event.eventType) {
                STTEventType.PARTIAL_RESULT -> {
                    event.transcript?.let { partial ->
                        partialTextView.text = partial
                    }
                }
                STTEventType.TRANSCRIPTION_COMPLETED -> {
                    event.transcript?.let { final ->
                        finalTextView.text = final
                    }
                }
                STTEventType.TRANSCRIPTION_FAILED -> {
                    showError(event.error ?: "Transcription failed")
                }
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/stt/stream#example-voice-search)

Example: Voice Search
------------------------------------------------------------------------------------------------

Copy

Ask AI

    class VoiceSearchActivity : AppCompatActivity() {
        private lateinit var audioRecorder: AudioRecorder
    
        override fun onCreate(savedInstanceState: Bundle?) {
            super.onCreate(savedInstanceState)
    
            // Load STT model
            lifecycleScope.launch {
                RunAnywhere.loadSTTModel("whisper-tiny")
            }
    
            // Subscribe to transcription events
            lifecycleScope.launch {
                RunAnywhere.events.sttEvents.collect { event ->
                    if (event.eventType == STTEventType.PARTIAL_RESULT) {
                        searchInput.setText(event.transcript)
                    }
                }
            }
    
            micButton.setOnClickListener {
                if (isRecording) {
                    stopRecording()
                } else {
                    startRecording()
                }
            }
        }
    
        private fun startRecording() {
            audioRecorder.start { samples ->
                lifecycleScope.launch {
                    RunAnywhere.processStreamingAudio(samples)
                }
            }
        }
    
        private fun stopRecording() {
            audioRecorder.stop()
            lifecycleScope.launch {
                RunAnywhere.stopStreamingTranscription()
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/stt/stream#performance-tips)

Performance Tips
---------------------------------------------------------------------------------------

For optimal streaming performance: - Process audio in chunks of 30-100ms - Use 16kHz sample rate for Whisper models - Consider using VAD to detect speech segments

| Use Case | Recommended Approach |
| --- | --- |
| Pre-recorded audio | `transcribe()` |
| Live microphone | `processStreamingAudio()` |
| Real-time dictation | Streaming + VAD |

[transcribe()](https://docs.runanywhere.ai/kotlin/stt/transcribe)
[STT Options](https://docs.runanywhere.ai/kotlin/stt/options)

⌘I

---

# Voice Activity Detection - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/vad#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Kotlin SDK

Voice Activity Detection

On this page

*   [Basic Usage](https://docs.runanywhere.ai/kotlin/vad#basic-usage)
    
*   [VADResult](https://docs.runanywhere.ai/kotlin/vad#vadresult)
    
*   [Configure VAD](https://docs.runanywhere.ai/kotlin/vad#configure-vad)
    
*   [VADConfiguration](https://docs.runanywhere.ai/kotlin/vad#vadconfiguration)
    
*   [Streaming VAD](https://docs.runanywhere.ai/kotlin/vad#streaming-vad)
    
*   [Calibrate with Ambient Noise](https://docs.runanywhere.ai/kotlin/vad#calibrate-with-ambient-noise)
    
*   [Example: Voice Recording with VAD](https://docs.runanywhere.ai/kotlin/vad#example-voice-recording-with-vad)
    
*   [VAD Statistics](https://docs.runanywhere.ai/kotlin/vad#vad-statistics)
    
*   [Reset VAD](https://docs.runanywhere.ai/kotlin/vad#reset-vad)
    
*   [Best Practices](https://docs.runanywhere.ai/kotlin/vad#best-practices)
    

Voice Activity Detection (VAD) identifies speech segments in audio streams, enabling responsive voice interfaces.

[​](https://docs.runanywhere.ai/kotlin/vad#basic-usage)

Basic Usage
----------------------------------------------------------------------

Copy

Ask AI

    // Detect speech in audio data
    val audioData: ByteArray = // ... from recording
    val result = RunAnywhere.detectVoiceActivity(audioData)
    
    if (result.hasSpeech) {
        println("Speech detected!")
        println("Confidence: ${result.confidence}")
    }
    

[​](https://docs.runanywhere.ai/kotlin/vad#vadresult)

VADResult
------------------------------------------------------------------

| Property | Type | Description |
| --- | --- | --- |
| `hasSpeech` | `Boolean` | Whether speech was detected |
| `confidence` | `Float` | Detection confidence (0.0-1.0) |
| `speechStartMs` | `Long?` | Start time of speech segment |
| `speechEndMs` | `Long?` | End time of speech segment |
| `frameIndex` | `Int` | Audio frame index |
| `timestamp` | `Long` | Detection timestamp |

[​](https://docs.runanywhere.ai/kotlin/vad#configure-vad)

Configure VAD
--------------------------------------------------------------------------

Customize detection sensitivity:

Copy

Ask AI

    RunAnywhere.configureVAD(VADConfiguration(
        threshold = 0.5f,              // Detection threshold (0.0-1.0)
        minSpeechDurationMs = 250,     // Minimum speech duration
        minSilenceDurationMs = 300,    // Silence before speech end
        sampleRate = 16000,            // Audio sample rate
        frameSizeMs = 30               // Frame size for processing
    ))
    

[​](https://docs.runanywhere.ai/kotlin/vad#vadconfiguration)

VADConfiguration
--------------------------------------------------------------------------------

Copy

Ask AI

    data class VADConfiguration(
        val threshold: Float = 0.5f,
        val minSpeechDurationMs: Int = 250,
        val minSilenceDurationMs: Int = 300,
        val sampleRate: Int = 16000,
        val frameSizeMs: Int = 30
    )
    

| Parameter | Default | Description |
| --- | --- | --- |
| `threshold` | 0.5 | Higher = less sensitive (fewer false positives) |
| `minSpeechDurationMs` | 250 | Ignore speech shorter than this |
| `minSilenceDurationMs` | 300 | Silence needed to end speech |
| `sampleRate` | 16000 | Audio sample rate |
| `frameSizeMs` | 30  | Frame duration for processing |

[​](https://docs.runanywhere.ai/kotlin/vad#streaming-vad)

Streaming VAD
--------------------------------------------------------------------------

Process continuous audio with Kotlin Flows:

Copy

Ask AI

    // Create a flow of audio samples from microphone
    val audioSamplesFlow: Flow<FloatArray> = microphoneManager.audioFlow
    
    // Stream VAD results
    RunAnywhere.streamVAD(audioSamplesFlow)
        .collect { result ->
            if (result.hasSpeech) {
                updateUI(state = "Speaking...")
            } else {
                updateUI(state = "Listening...")
            }
        }
    

[​](https://docs.runanywhere.ai/kotlin/vad#calibrate-with-ambient-noise)

Calibrate with Ambient Noise
--------------------------------------------------------------------------------------------------------

Improve accuracy by calibrating with ambient noise:

Copy

Ask AI

    // Record a few seconds of ambient noise
    val ambientAudio = recordAmbientNoise(durationMs = 2000)
    
    // Calibrate VAD
    RunAnywhere.calibrateVAD(ambientAudio)
    

[​](https://docs.runanywhere.ai/kotlin/vad#example-voice-recording-with-vad)

Example: Voice Recording with VAD
-----------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class VoiceRecorderViewModel : ViewModel() {
        private val _isListening = MutableStateFlow(false)
        val isListening: StateFlow<Boolean> = _isListening
    
        private val _isSpeaking = MutableStateFlow(false)
        val isSpeaking: StateFlow<Boolean> = _isSpeaking
    
        private val audioBuffer = mutableListOf<FloatArray>()
    
        fun startListening(audioFlow: Flow<FloatArray>) {
            viewModelScope.launch {
                _isListening.value = true
    
                // Configure VAD
                RunAnywhere.configureVAD(VADConfiguration(
                    threshold = 0.5f,
                    minSpeechDurationMs = 300,
                    minSilenceDurationMs = 500
                ))
    
                // Stream VAD
                RunAnywhere.streamVAD(audioFlow)
                    .collect { result ->
                        _isSpeaking.value = result.hasSpeech
    
                        if (result.hasSpeech) {
                            // Buffer audio during speech
                            // (actual audio capture would be handled separately)
                        } else if (audioBuffer.isNotEmpty()) {
                            // Speech ended - process buffered audio
                            processRecordedAudio()
                        }
                    }
            }
        }
    
        fun stopListening() {
            viewModelScope.launch {
                _isListening.value = false
                RunAnywhere.resetVAD()
            }
        }
    
        private suspend fun processRecordedAudio() {
            // Convert buffer to ByteArray and transcribe
            val audioData = audioBuffer.toByteArray()
            val transcription = RunAnywhere.transcribe(audioData)
            _transcription.value = transcription
            audioBuffer.clear()
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/vad#vad-statistics)

VAD Statistics
----------------------------------------------------------------------------

Get current VAD performance metrics:

Copy

Ask AI

    val stats = RunAnywhere.getVADStatistics()
    println("Total frames processed: ${stats.framesProcessed}")
    println("Speech frames: ${stats.speechFrames}")
    println("Current state: ${stats.currentState}")
    

[​](https://docs.runanywhere.ai/kotlin/vad#reset-vad)

Reset VAD
------------------------------------------------------------------

Clear VAD state for a fresh start:

Copy

Ask AI

    RunAnywhere.resetVAD()
    

[​](https://docs.runanywhere.ai/kotlin/vad#best-practices)

Best Practices
----------------------------------------------------------------------------

**Calibration**: Always calibrate VAD with ambient noise for best results in noisy environments.**Threshold tuning**:

*   Quiet environment: threshold = 0.3-0.5
*   Noisy environment: threshold = 0.6-0.8

**Buffer management**: Use VAD to determine when to start/stop recording, but buffer audio slightly before speech detection for complete captures.

[Voices](https://docs.runanywhere.ai/kotlin/tts/voices)
[Voice Agent](https://docs.runanywhere.ai/kotlin/voice-agent)

⌘I

---

# transcribe() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/stt/transcribe#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

transcribe()

On this page

*   [Basic Transcription](https://docs.runanywhere.ai/kotlin/stt/transcribe#basic-transcription)
    
*   [Transcription with Options](https://docs.runanywhere.ai/kotlin/stt/transcribe#transcription-with-options)
    
*   [STTOutput](https://docs.runanywhere.ai/kotlin/stt/transcribe#sttoutput)
    
*   [Example: Transcribe Audio File](https://docs.runanywhere.ai/kotlin/stt/transcribe#example-transcribe-audio-file)
    
*   [Example: Record and Transcribe](https://docs.runanywhere.ai/kotlin/stt/transcribe#example-record-and-transcribe)
    
*   [Model Management](https://docs.runanywhere.ai/kotlin/stt/transcribe#model-management)
    
*   [Supported Audio Formats](https://docs.runanywhere.ai/kotlin/stt/transcribe#supported-audio-formats)
    

Transcribe audio data to text using on-device speech recognition models.

[​](https://docs.runanywhere.ai/kotlin/stt/transcribe#basic-transcription)

Basic Transcription
-------------------------------------------------------------------------------------------------

Copy

Ask AI

    // Load an STT model first
    RunAnywhere.loadSTTModel("whisper-tiny")
    
    // Transcribe audio bytes
    val audioData: ByteArray = // ... from file or recording
    val text = RunAnywhere.transcribe(audioData)
    println(text)  // "Hello, how are you today?"
    

[​](https://docs.runanywhere.ai/kotlin/stt/transcribe#transcription-with-options)

Transcription with Options
---------------------------------------------------------------------------------------------------------------

Get detailed output including confidence scores and timestamps:

Copy

Ask AI

    val output = RunAnywhere.transcribeWithOptions(
        audioData = audioBytes,
        options = STTOptions(
            language = "en",
            enableTimestamps = true,
            enablePunctuation = true
        )
    )
    
    println("Text: ${output.text}")
    println("Confidence: ${output.confidence}")
    
    // Access word-level timestamps
    output.wordTimestamps?.forEach { word ->
        println("[${word.startTime}s - ${word.endTime}s]: ${word.word}")
    }
    

[​](https://docs.runanywhere.ai/kotlin/stt/transcribe#sttoutput)

STTOutput
-----------------------------------------------------------------------------

The detailed output object:

| Property | Type | Description |
| --- | --- | --- |
| `text` | `String` | Transcribed text |
| `confidence` | `Float` | Confidence score (0.0-1.0) |
| `wordTimestamps` | `List<WordTimestamp>?` | Word-level timing |
| `detectedLanguage` | `String?` | Auto-detected language code |
| `metadata` | `TranscriptionMetadata` | Processing metrics |

[​](https://docs.runanywhere.ai/kotlin/stt/transcribe#example-transcribe-audio-file)

Example: Transcribe Audio File
----------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    suspend fun transcribeAudioFile(uri: Uri): String {
        val audioData = contentResolver.openInputStream(uri)?.readBytes()
            ?: throw IllegalArgumentException("Cannot read audio file")
    
        // Ensure STT model is loaded
        if (!RunAnywhere.isSTTModelLoaded()) {
            RunAnywhere.loadSTTModel("whisper-tiny")
        }
    
        return RunAnywhere.transcribe(audioData)
    }
    

[​](https://docs.runanywhere.ai/kotlin/stt/transcribe#example-record-and-transcribe)

Example: Record and Transcribe
----------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class VoiceRecorderViewModel : ViewModel() {
        private var audioRecorder: AudioRecord? = null
        private val audioBuffer = mutableListOf<Byte>()
    
        fun startRecording() {
            // Start recording audio...
        }
    
        fun stopAndTranscribe() {
            viewModelScope.launch {
                val audioData = audioBuffer.toByteArray()
    
                val result = RunAnywhere.transcribeWithOptions(
                    audioData,
                    options = STTOptions(
                        language = "en",
                        enablePunctuation = true
                    )
                )
    
                _transcription.value = result.text
                _confidence.value = result.confidence
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/stt/transcribe#model-management)

Model Management
-------------------------------------------------------------------------------------------

Copy

Ask AI

    // Load a specific STT model
    RunAnywhere.loadSTTModel("whisper-tiny")
    
    // Check if loaded
    val isLoaded = RunAnywhere.isSTTModelLoaded()
    
    // Get current model ID
    val modelId = RunAnywhere.currentSTTModelId
    
    // Unload when done
    RunAnywhere.unloadSTTModel()
    

[​](https://docs.runanywhere.ai/kotlin/stt/transcribe#supported-audio-formats)

Supported Audio Formats
---------------------------------------------------------------------------------------------------------

| Format | Sample Rate | Notes |
| --- | --- | --- |
| PCM | 16000 Hz | Recommended for best quality |
| WAV | 16000 Hz | Standard audio file format |
| MP3 | Any | Converted internally |

For best transcription accuracy: - Use 16kHz mono PCM audio - Keep audio clips under 30 seconds for optimal performance - Use a smaller model (whisper-tiny) for faster results, larger models (whisper-base) for better accuracy

[System Prompts](https://docs.runanywhere.ai/kotlin/llm/system-prompts)
[Streaming STT](https://docs.runanywhere.ai/kotlin/stt/stream)

⌘I

---

# generateStream() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/llm/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

generateStream()

On this page

*   [Basic Streaming](https://docs.runanywhere.ai/kotlin/llm/stream#basic-streaming)
    
*   [Streaming with Metrics](https://docs.runanywhere.ai/kotlin/llm/stream#streaming-with-metrics)
    
*   [LLMStreamingResult](https://docs.runanywhere.ai/kotlin/llm/stream#llmstreamingresult)
    
*   [Example: Chat UI with Streaming](https://docs.runanywhere.ai/kotlin/llm/stream#example-chat-ui-with-streaming)
    
*   [Example: Jetpack Compose Integration](https://docs.runanywhere.ai/kotlin/llm/stream#example-jetpack-compose-integration)
    
*   [Performance Tips](https://docs.runanywhere.ai/kotlin/llm/stream#performance-tips)
    

Use streaming for responsive UIs that display text as it’s generated, providing a better user experience for longer responses.

[​](https://docs.runanywhere.ai/kotlin/llm/stream#basic-streaming)

Basic Streaming
-------------------------------------------------------------------------------------

Copy

Ask AI

    RunAnywhere.generateStream("Tell me a story about AI")
        .collect { token ->
            // Display each token as it arrives
            print(token)
            textView.append(token)
        }
    

[​](https://docs.runanywhere.ai/kotlin/llm/stream#streaming-with-metrics)

Streaming with Metrics
---------------------------------------------------------------------------------------------------

Get both the token stream AND final metrics:

Copy

Ask AI

    val streamResult = RunAnywhere.generateStreamWithMetrics(
        prompt = "Explain quantum computing",
        options = LLMGenerationOptions(maxTokens = 500)
    )
    
    // Collect tokens as they arrive
    streamResult.stream.collect { token ->
        textView.append(token)
    }
    
    // Get final metrics after streaming completes
    val metrics = streamResult.result.await()
    println("\n\nGenerated ${metrics.tokensUsed} tokens")
    println("Speed: ${metrics.tokensPerSecond} tok/s")
    println("Time to first token: ${metrics.timeToFirstTokenMs}ms")
    

[​](https://docs.runanywhere.ai/kotlin/llm/stream#llmstreamingresult)

LLMStreamingResult
-------------------------------------------------------------------------------------------

Copy

Ask AI

    data class LLMStreamingResult(
        val stream: Flow<String>,                    // Token stream
        val result: Deferred<LLMGenerationResult>    // Final metrics (awaitable)
    )
    

[​](https://docs.runanywhere.ai/kotlin/llm/stream#example-chat-ui-with-streaming)

Example: Chat UI with Streaming
--------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class ChatViewModel : ViewModel() {
        private val _currentResponse = MutableStateFlow("")
        val currentResponse: StateFlow<String> = _currentResponse
    
        private val _isGenerating = MutableStateFlow(false)
        val isGenerating: StateFlow<Boolean> = _isGenerating
    
        fun sendMessage(prompt: String) {
            viewModelScope.launch {
                _isGenerating.value = true
                _currentResponse.value = ""
    
                try {
                    RunAnywhere.generateStream(prompt)
                        .collect { token ->
                            _currentResponse.value += token
                        }
                } catch (e: Exception) {
                    _currentResponse.value = "Error: ${e.message}"
                } finally {
                    _isGenerating.value = false
                }
            }
        }
    
        fun cancelGeneration() {
            RunAnywhere.cancelGeneration()
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/llm/stream#example-jetpack-compose-integration)

Example: Jetpack Compose Integration
------------------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    @Composable
    fun StreamingTextView(viewModel: ChatViewModel) {
        val response by viewModel.currentResponse.collectAsState()
        val isGenerating by viewModel.isGenerating.collectAsState()
    
        Column {
            Text(
                text = response,
                modifier = Modifier
                    .fillMaxWidth()
                    .padding(16.dp)
            )
    
            if (isGenerating) {
                LinearProgressIndicator(
                    modifier = Modifier.fillMaxWidth()
                )
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/llm/stream#performance-tips)

Performance Tips
---------------------------------------------------------------------------------------

Streaming provides better perceived performance for long generations: - **Time to First Token (TTFT)**: ~50-100ms - Users see immediate feedback instead of waiting for the full response

| Scenario | Recommended Approach |
| --- | --- |
| Short responses (under 50 tokens) | `generate()` |
| Long responses (over 100 tokens) | `generateStream()` |
| Chat interfaces | `generateStream()` |
| Background processing | `generate()` |

[generate()](https://docs.runanywhere.ai/kotlin/llm/generate)
[System Prompts](https://docs.runanywhere.ai/kotlin/llm/system-prompts)

⌘I

---

# Voices - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/tts/voices#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

Voices

On this page

*   [List Available Voices](https://docs.runanywhere.ai/kotlin/tts/voices#list-available-voices)
    
*   [Load a Voice](https://docs.runanywhere.ai/kotlin/tts/voices#load-a-voice)
    
*   [Voice Configuration](https://docs.runanywhere.ai/kotlin/tts/voices#voice-configuration)
    
*   [Rate, Pitch, and Volume](https://docs.runanywhere.ai/kotlin/tts/voices#rate-pitch-and-volume)
    
*   [Example: Voice Selector](https://docs.runanywhere.ai/kotlin/tts/voices#example-voice-selector)
    
*   [Sherpa-ONNX Voices](https://docs.runanywhere.ai/kotlin/tts/voices#sherpa-onnx-voices)
    
*   [Download Additional Voices](https://docs.runanywhere.ai/kotlin/tts/voices#download-additional-voices)
    
*   [Best Practices](https://docs.runanywhere.ai/kotlin/tts/voices#best-practices)
    

[​](https://docs.runanywhere.ai/kotlin/tts/voices#list-available-voices)

List Available Voices
-------------------------------------------------------------------------------------------------

Copy

Ask AI

    val voices = RunAnywhere.availableTTSVoices()
    voices.forEach { voiceId ->
        println(voiceId)
    }
    

[​](https://docs.runanywhere.ai/kotlin/tts/voices#load-a-voice)

Load a Voice
-------------------------------------------------------------------------------

Copy

Ask AI

    // Load a specific voice
    RunAnywhere.loadTTSVoice("en-us-default")
    
    // Check if voice is loaded
    val isLoaded = RunAnywhere.isTTSVoiceLoaded()
    
    // Get current voice ID
    val currentVoice = RunAnywhere.currentTTSVoiceId
    
    // Unload when done
    RunAnywhere.unloadTTSVoice()
    

[​](https://docs.runanywhere.ai/kotlin/tts/voices#voice-configuration)

Voice Configuration
---------------------------------------------------------------------------------------------

Customize voice parameters in TTSOptions:

Copy

Ask AI

    val options = TTSOptions(
        voice = "en-US-Neural2-F",   // Specific voice ID
        rate = 0.8f,                  // 0.0-2.0, slower speed
        pitch = 1.3f,                 // 0.0-2.0, higher pitch
        volume = 1.0f                 // 0.0-1.0
    )
    
    val output = RunAnywhere.synthesize("Hello world", options)
    

[​](https://docs.runanywhere.ai/kotlin/tts/voices#rate-pitch-and-volume)

Rate, Pitch, and Volume
---------------------------------------------------------------------------------------------------

| Parameter | Range | Default | Effect |
| --- | --- | --- | --- |
| `rate` | 0.0-2.0 | 1.0 | Speech speed (0.5 = half speed, 2.0 = double speed) |
| `pitch` | 0.0-2.0 | 1.0 | Voice pitch (lower = deeper, higher = lighter) |
| `volume` | 0.0-1.0 | 1.0 | Audio volume |

[​](https://docs.runanywhere.ai/kotlin/tts/voices#example-voice-selector)

Example: Voice Selector
----------------------------------------------------------------------------------------------------

Copy

Ask AI

    class VoiceSelectorViewModel : ViewModel() {
        private val _availableVoices = MutableStateFlow<List<String>>(emptyList())
        val availableVoices: StateFlow<List<String>> = _availableVoices
    
        private val _selectedVoice = MutableStateFlow<String?>(null)
        val selectedVoice: StateFlow<String?> = _selectedVoice
    
        init {
            viewModelScope.launch {
                _availableVoices.value = RunAnywhere.availableTTSVoices()
            }
        }
    
        fun selectVoice(voiceId: String) {
            viewModelScope.launch {
                RunAnywhere.loadTTSVoice(voiceId)
                _selectedVoice.value = voiceId
            }
        }
    
        fun previewVoice(voiceId: String) {
            viewModelScope.launch {
                RunAnywhere.loadTTSVoice(voiceId)
                RunAnywhere.speak(
                    "Hello! This is a preview of the selected voice.",
                    TTSOptions(rate = 1.0f)
                )
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/tts/voices#sherpa-onnx-voices)

Sherpa-ONNX Voices
-------------------------------------------------------------------------------------------

The SDK uses Sherpa-ONNX for neural TTS. Available voice packs include:

| Voice ID | Language | Description |
| --- | --- | --- |
| `en-us-default` | English (US) | Default English voice |
| `en-gb-default` | English (UK) | British English voice |
| `de-de-default` | German | German voice |
| `fr-fr-default` | French | French voice |
| `es-es-default` | Spanish | Spanish voice |

Voice availability depends on downloaded voice models. Use `availableTTSVoices()` to see what’s available on the device.

[​](https://docs.runanywhere.ai/kotlin/tts/voices#download-additional-voices)

Download Additional Voices
-----------------------------------------------------------------------------------------------------------

Register and download additional TTS voice models:

Copy

Ask AI

    // Register a TTS voice model
    val voiceModel = RunAnywhere.registerModel(
        name = "Piper English Amy",
        url = "https://example.com/piper-en-us-amy.onnx",
        framework = InferenceFramework.ONNX,
        modality = ModelCategory.SPEECH_SYNTHESIS
    )
    
    // Download the voice
    RunAnywhere.downloadModel(voiceModel.id).collect { progress ->
        println("Downloading: ${(progress.progress * 100).toInt()}%")
    }
    
    // Load and use
    RunAnywhere.loadTTSVoice(voiceModel.id)
    

[​](https://docs.runanywhere.ai/kotlin/tts/voices#best-practices)

Best Practices
-----------------------------------------------------------------------------------

*   **Preload voices** at app startup for instant TTS - **Use streaming** for long text to reduce latency - **Cache common phrases** as audio files for fastest playback - **Match voice language** to content language for best quality

[Streaming TTS](https://docs.runanywhere.ai/kotlin/tts/stream)
[Voice Activity Detection](https://docs.runanywhere.ai/kotlin/vad)

⌘I

---

# STT Options - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/stt/options#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

STT Options

On this page

*   [STTOptions](https://docs.runanywhere.ai/kotlin/stt/options#sttoptions)
    
*   [Configuration Examples](https://docs.runanywhere.ai/kotlin/stt/options#configuration-examples)
    
*   [Basic English Transcription](https://docs.runanywhere.ai/kotlin/stt/options#basic-english-transcription)
    
*   [Multi-Language Detection](https://docs.runanywhere.ai/kotlin/stt/options#multi-language-detection)
    
*   [Transcription with Timestamps](https://docs.runanywhere.ai/kotlin/stt/options#transcription-with-timestamps)
    
*   [Speaker Diarization](https://docs.runanywhere.ai/kotlin/stt/options#speaker-diarization)
    
*   [Custom Vocabulary](https://docs.runanywhere.ai/kotlin/stt/options#custom-vocabulary)
    
*   [Supported Languages](https://docs.runanywhere.ai/kotlin/stt/options#supported-languages)
    
*   [AudioFormat](https://docs.runanywhere.ai/kotlin/stt/options#audioformat)
    
*   [TranscriptionMetadata](https://docs.runanywhere.ai/kotlin/stt/options#transcriptionmetadata)
    

Customize transcription behavior with `STTOptions`.

[​](https://docs.runanywhere.ai/kotlin/stt/options#sttoptions)

STTOptions
----------------------------------------------------------------------------

Copy

Ask AI

    data class STTOptions(
        val language: String = "en",              // BCP-47 language code
        val detectLanguage: Boolean = false,      // Auto-detect spoken language
        val enablePunctuation: Boolean = true,    // Add punctuation
        val enableDiarization: Boolean = false,   // Identify speakers
        val maxSpeakers: Int? = null,             // Max speakers to identify
        val enableTimestamps: Boolean = true,     // Word-level timestamps
        val vocabularyFilter: List<String> = emptyList(),  // Custom vocabulary
        val audioFormat: AudioFormat = AudioFormat.PCM,
        val sampleRate: Int = 16000               // Audio sample rate
    )
    

[​](https://docs.runanywhere.ai/kotlin/stt/options#configuration-examples)

Configuration Examples
----------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/kotlin/stt/options#basic-english-transcription)

Basic English Transcription

Copy

Ask AI

    val options = STTOptions(
        language = "en",
        enablePunctuation = true
    )
    

### 

[​](https://docs.runanywhere.ai/kotlin/stt/options#multi-language-detection)

Multi-Language Detection

Copy

Ask AI

    val options = STTOptions(
        detectLanguage = true,
        enablePunctuation = true
    )
    
    val result = RunAnywhere.transcribeWithOptions(audioData, options)
    println("Detected language: ${result.detectedLanguage}")
    

### 

[​](https://docs.runanywhere.ai/kotlin/stt/options#transcription-with-timestamps)

Transcription with Timestamps

Copy

Ask AI

    val options = STTOptions(
        language = "en",
        enableTimestamps = true
    )
    
    val result = RunAnywhere.transcribeWithOptions(audioData, options)
    result.wordTimestamps?.forEach { word ->
        println("[${word.startTime}s - ${word.endTime}s]: ${word.word}")
    }
    

### 

[​](https://docs.runanywhere.ai/kotlin/stt/options#speaker-diarization)

Speaker Diarization

Copy

Ask AI

    val options = STTOptions(
        language = "en",
        enableDiarization = true,
        maxSpeakers = 2
    )
    

### 

[​](https://docs.runanywhere.ai/kotlin/stt/options#custom-vocabulary)

Custom Vocabulary

Improve recognition for domain-specific terms:

Copy

Ask AI

    val options = STTOptions(
        language = "en",
        vocabularyFilter = listOf(
            "RunAnywhere",
            "Kotlin",
            "ONNX",
            "llama.cpp"
        )
    )
    

[​](https://docs.runanywhere.ai/kotlin/stt/options#supported-languages)

Supported Languages
----------------------------------------------------------------------------------------------

| Code | Language |
| --- | --- |
| `en` | English |
| `es` | Spanish |
| `fr` | French |
| `de` | German |
| `it` | Italian |
| `pt` | Portuguese |
| `zh` | Chinese |
| `ja` | Japanese |
| `ko` | Korean |

Language support depends on the STT model. Whisper models support 99+ languages.

[​](https://docs.runanywhere.ai/kotlin/stt/options#audioformat)

AudioFormat
------------------------------------------------------------------------------

Copy

Ask AI

    enum class AudioFormat {
        PCM,    // Raw PCM samples
        WAV,    // WAV container
        MP3,    // MP3 compressed
        AAC,    // AAC compressed
        OGG,    // Ogg Vorbis
        OPUS,   // Opus codec
        FLAC    // FLAC lossless
    }
    

[​](https://docs.runanywhere.ai/kotlin/stt/options#transcriptionmetadata)

TranscriptionMetadata
--------------------------------------------------------------------------------------------------

Access processing metrics:

Copy

Ask AI

    val result = RunAnywhere.transcribeWithOptions(audioData, options)
    
    val metadata = result.metadata
    println("Model: ${metadata.modelId}")
    println("Processing time: ${metadata.processingTime}s")
    println("Audio length: ${metadata.audioLength}s")
    println("Real-time factor: ${metadata.realTimeFactor}x")
    

| Property | Description |
| --- | --- |
| `modelId` | STT model used |
| `processingTime` | Time to process (seconds) |
| `audioLength` | Input audio duration (seconds) |
| `realTimeFactor` | Processing time / audio length |

A real-time factor < 1.0 means the transcription is faster than real-time. For example, 0.2x means 10 seconds of audio processed in 2 seconds.

[Streaming STT](https://docs.runanywhere.ai/kotlin/stt/stream)
[synthesize()](https://docs.runanywhere.ai/kotlin/tts/synthesize)

⌘I

---

# Quick Start - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/quick-start#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Kotlin SDK

Quick Start

On this page

*   [Complete Example](https://docs.runanywhere.ai/kotlin/quick-start#complete-example)
    
*   [Step-by-Step Guide](https://docs.runanywhere.ai/kotlin/quick-start#step-by-step-guide)
    
*   [1\. Initialize the SDK](https://docs.runanywhere.ai/kotlin/quick-start#1-initialize-the-sdk)
    
*   [Environment Options](https://docs.runanywhere.ai/kotlin/quick-start#environment-options)
    
*   [2\. Register a Model](https://docs.runanywhere.ai/kotlin/quick-start#2-register-a-model)
    
*   [3\. Download the Model](https://docs.runanywhere.ai/kotlin/quick-start#3-download-the-model)
    
*   [4\. Load and Generate](https://docs.runanywhere.ai/kotlin/quick-start#4-load-and-generate)
    
*   [5\. Stream Responses](https://docs.runanywhere.ai/kotlin/quick-start#5-stream-responses)
    
*   [What’s Next?](https://docs.runanywhere.ai/kotlin/quick-start#what%E2%80%99s-next)
    

[​](https://docs.runanywhere.ai/kotlin/quick-start#complete-example)

Complete Example
----------------------------------------------------------------------------------------

Here’s a complete example to get you started with on-device text generation:

MainActivity.kt

Copy

Ask AI

    import com.runanywhere.sdk.public.RunAnywhere
    import com.runanywhere.sdk.public.SDKEnvironment
    import com.runanywhere.sdk.public.extensions.*
    import com.runanywhere.sdk.core.types.InferenceFramework
    
    class MyApplication : Application() {
        override fun onCreate() {
            super.onCreate()
    
            // 1. Initialize SDK (fast, ~1-5ms)
            RunAnywhere.initialize(
                apiKey = "your-api-key",    // Optional for development
                environment = SDKEnvironment.DEVELOPMENT
            )
        }
    }
    
    // 2. Use in your Activity/Fragment
    class MainActivity : AppCompatActivity() {
        private var modelId: String? = null
    
        override fun onCreate(savedInstanceState: Bundle?) {
            super.onCreate(savedInstanceState)
    
            lifecycleScope.launch {
                // Register and download a model
                val modelInfo = RunAnywhere.registerModel(
                    name = "Qwen 0.5B",
                    url = "https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF/resolve/main/qwen2.5-0.5b-instruct-q8_0.gguf",
                    framework = InferenceFramework.LLAMA_CPP
                )
                modelId = modelInfo.id
    
                // Download with progress tracking
                RunAnywhere.downloadModel(modelInfo.id).collect { progress ->
                    updateProgress((progress.progress * 100).toInt())
                }
    
                // Load and generate
                RunAnywhere.loadLLMModel(modelInfo.id)
    
                val result = RunAnywhere.generate(
                    prompt = "Explain quantum computing in simple terms",
                    options = LLMGenerationOptions(
                        maxTokens = 200,
                        temperature = 0.7f
                    )
                )
    
                showResponse(result.text)
                Log.d("LLM", "Generated in ${result.latencyMs}ms at ${result.tokensPerSecond} tok/s")
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/quick-start#step-by-step-guide)

Step-by-Step Guide
--------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/kotlin/quick-start#1-initialize-the-sdk)

1\. Initialize the SDK

Initialize RunAnywhere once in your `Application.onCreate()`:

Copy

Ask AI

    RunAnywhere.initialize(
        apiKey = "your-api-key",       // Optional for development mode
        baseURL = null,                // Uses default API endpoint
        environment = SDKEnvironment.DEVELOPMENT
    )
    

### 

[​](https://docs.runanywhere.ai/kotlin/quick-start#environment-options)

Environment Options

| Environment | Log Level | Description |
| --- | --- | --- |
| `DEVELOPMENT` | Debug | Full logging, local testing |
| `STAGING` | Info | Staging backend, moderate logging |
| `PRODUCTION` | Warning | Production backend, minimal logging |

### 

[​](https://docs.runanywhere.ai/kotlin/quick-start#2-register-a-model)

2\. Register a Model

Register models from HuggingFace or any direct URL:

Copy

Ask AI

    val modelInfo = RunAnywhere.registerModel(
        name = "Qwen 0.5B",
        url = "https://huggingface.co/.../model.gguf",
        framework = InferenceFramework.LLAMA_CPP
    )
    

### 

[​](https://docs.runanywhere.ai/kotlin/quick-start#3-download-the-model)

3\. Download the Model

Download with progress tracking using Kotlin Flows:

Copy

Ask AI

    RunAnywhere.downloadModel(modelInfo.id).collect { progress ->
        when (progress.state) {
            DownloadState.DOWNLOADING -> {
                val percent = (progress.progress * 100).toInt()
                progressBar.progress = percent
            }
            DownloadState.COMPLETED -> {
                showMessage("Download complete!")
            }
            DownloadState.ERROR -> {
                showError(progress.error ?: "Download failed")
            }
        }
    }
    

### 

[​](https://docs.runanywhere.ai/kotlin/quick-start#4-load-and-generate)

4\. Load and Generate

Copy

Ask AI

    // Load the model into memory
    RunAnywhere.loadLLMModel(modelInfo.id)
    
    // Simple chat
    val response = RunAnywhere.chat("What is 2+2?")
    
    // Or with full metrics
    val result = RunAnywhere.generate(
        prompt = "Write a haiku about coding",
        options = LLMGenerationOptions(maxTokens = 50)
    )
    println("Response: ${result.text}")
    println("Tokens/sec: ${result.tokensPerSecond}")
    

### 

[​](https://docs.runanywhere.ai/kotlin/quick-start#5-stream-responses)

5\. Stream Responses

For real-time token streaming:

Copy

Ask AI

    RunAnywhere.generateStream("Tell me a story about AI")
        .collect { token ->
            // Display tokens as they arrive
            textView.append(token)
        }
    

[​](https://docs.runanywhere.ai/kotlin/quick-start#what%E2%80%99s-next)

What’s Next?
---------------------------------------------------------------------------------------

[LLM Chat\
--------\
\
Simple one-liner for quick responses](https://docs.runanywhere.ai/kotlin/llm/chat)
[Streaming\
---------\
\
Real-time token streaming with Flows](https://docs.runanywhere.ai/kotlin/llm/stream)
[Speech-to-Text\
--------------\
\
Transcribe audio to text](https://docs.runanywhere.ai/kotlin/stt/transcribe)
[Text-to-Speech\
--------------\
\
Convert text to spoken audio](https://docs.runanywhere.ai/kotlin/tts/synthesize)

[Installation](https://docs.runanywhere.ai/kotlin/installation)
[chat()](https://docs.runanywhere.ai/kotlin/llm/chat)

⌘I

---

# Installation - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/installation#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

React Native SDK

Installation

On this page

*   [Package Installation](https://docs.runanywhere.ai/react-native/installation#package-installation)
    
*   [Full Installation (All Features)](https://docs.runanywhere.ai/react-native/installation#full-installation-all-features)
    
*   [LLM Only (Text Generation)](https://docs.runanywhere.ai/react-native/installation#llm-only-text-generation)
    
*   [Voice Only (STT/TTS)](https://docs.runanywhere.ai/react-native/installation#voice-only-stt%2Ftts)
    
*   [Required Peer Dependencies](https://docs.runanywhere.ai/react-native/installation#required-peer-dependencies)
    
*   [Package Overview](https://docs.runanywhere.ai/react-native/installation#package-overview)
    
*   [Supported Model Formats](https://docs.runanywhere.ai/react-native/installation#supported-model-formats)
    
*   [iOS Setup](https://docs.runanywhere.ai/react-native/installation#ios-setup)
    
*   [iOS Permissions](https://docs.runanywhere.ai/react-native/installation#ios-permissions)
    
*   [Android Setup](https://docs.runanywhere.ai/react-native/installation#android-setup)
    
*   [Android Build Requirements](https://docs.runanywhere.ai/react-native/installation#android-build-requirements)
    
*   [Android Permissions](https://docs.runanywhere.ai/react-native/installation#android-permissions)
    
*   [Running on Physical Device](https://docs.runanywhere.ai/react-native/installation#running-on-physical-device)
    
*   [ProGuard Rules](https://docs.runanywhere.ai/react-native/installation#proguard-rules)
    
*   [Troubleshooting Installation](https://docs.runanywhere.ai/react-native/installation#troubleshooting-installation)
    
*   [iOS: Pod Install Fails](https://docs.runanywhere.ai/react-native/installation#ios-pod-install-fails)
    
*   [Android: Native Library Not Found](https://docs.runanywhere.ai/react-native/installation#android-native-library-not-found)
    
*   [Android: “hermesEnabled” Property Not Found](https://docs.runanywhere.ai/react-native/installation#android-%E2%80%9Chermesenabled%E2%80%9D-property-not-found)
    
*   [Android: “Unable to load script” on Device](https://docs.runanywhere.ai/react-native/installation#android-%E2%80%9Cunable-to-load-script%E2%80%9D-on-device)
    
*   [Android: Grey Screen After Launch](https://docs.runanywhere.ai/react-native/installation#android-grey-screen-after-launch)
    
*   [Metro: Module Resolution Issues](https://docs.runanywhere.ai/react-native/installation#metro-module-resolution-issues)
    
*   [Both Platforms: Fresh Build](https://docs.runanywhere.ai/react-native/installation#both-platforms-fresh-build)
    
*   [Native Binary Sizes](https://docs.runanywhere.ai/react-native/installation#native-binary-sizes)
    
*   [Next Steps](https://docs.runanywhere.ai/react-native/installation#next-steps)
    

[​](https://docs.runanywhere.ai/react-native/installation#package-installation)

Package Installation
-------------------------------------------------------------------------------------------------------

Choose the packages based on your feature requirements:

### 

[​](https://docs.runanywhere.ai/react-native/installation#full-installation-all-features)

Full Installation (All Features)

Copy

Ask AI

    npm install @runanywhere/core @runanywhere/llamacpp @runanywhere/onnx
    # or
    yarn add @runanywhere/core @runanywhere/llamacpp @runanywhere/onnx
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#llm-only-text-generation)

LLM Only (Text Generation)

Copy

Ask AI

    npm install @runanywhere/core @runanywhere/llamacpp
    # or
    yarn add @runanywhere/core @runanywhere/llamacpp
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#voice-only-stt/tts)

Voice Only (STT/TTS)

Copy

Ask AI

    npm install @runanywhere/core @runanywhere/onnx
    # or
    yarn add @runanywhere/core @runanywhere/onnx
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#required-peer-dependencies)

Required Peer Dependencies

The SDK uses Nitro Modules for its native bridge and requires file system access for model downloads:

Copy

Ask AI

    # Required: Native bridge layer
    npm install react-native-nitro-modules
    
    # Required: File system access for model downloads
    npm install react-native-fs@^2.20.0 react-native-zip-archive@^6.1.0
    

`react-native-fs` and `react-native-zip-archive` are **required** for model downloads. Without them, you will see: `ERROR: react-native-fs not installed - cannot download models`

Include only the SDK modules you need. Each native module adds to your app size: - **@runanywhere/core**: ~2MB (required) - **@runanywhere/llamacpp**: ~15-25MB (for LLM text generation) - **@runanywhere/onnx**: ~50-70MB (for STT, TTS, and VAD)

[​](https://docs.runanywhere.ai/react-native/installation#package-overview)

Package Overview
-----------------------------------------------------------------------------------------------

| Package | Size | Features |
| --- | --- | --- |
| `@runanywhere/core` | ~2MB | Core SDK, model management, events |
| `@runanywhere/llamacpp` | ~15-25MB | LLM text generation (GGUF models) |
| `@runanywhere/onnx` | ~50-70MB | STT, TTS, VAD (ONNX models) |

[​](https://docs.runanywhere.ai/react-native/installation#supported-model-formats)

Supported Model Formats
-------------------------------------------------------------------------------------------------------------

| Format | Extension | Backend | Use Case |
| --- | --- | --- | --- |
| GGUF | `.gguf` | LlamaCPP | LLM text generation |
| ONNX | `.onnx` | ONNX Runtime | STT, TTS, VAD |

[​](https://docs.runanywhere.ai/react-native/installation#ios-setup)

iOS Setup
---------------------------------------------------------------------------------

After installing packages, run CocoaPods:

Copy

Ask AI

    cd ios && pod install && cd ..
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#ios-permissions)

iOS Permissions

Add required permissions to your `Info.plist`:

Info.plist

Copy

Ask AI

    <!-- Required for model downloads -->
    <key>NSAppTransportSecurity</key>
    <dict>
      <key>NSAllowsArbitraryLoads</key>
      <true/>
    </dict>
    
    <!-- Required for STT/VAD with microphone -->
    <key>NSMicrophoneUsageDescription</key>
    <string>This app needs microphone access for voice features.</string>
    
    <!-- Required for background audio (optional) -->
    <key>UIBackgroundModes</key>
    <array>
      <string>audio</string>
    </array>
    

`NSMicrophoneUsageDescription` is only required if you’re using STT or VAD with microphone input.

[​](https://docs.runanywhere.ai/react-native/installation#android-setup)

Android Setup
-----------------------------------------------------------------------------------------

Native libraries are automatically downloaded during the Gradle build.

### 

[​](https://docs.runanywhere.ai/react-native/installation#android-build-requirements)

Android Build Requirements

| Requirement | Minimum Version |
| --- | --- |
| Android SDK | compileSdk 36 |
| Android NDK | 27.1.12297006 |
| JDK | 17+ |
| Min API | 24 (Android 7.0+) |

### 

[​](https://docs.runanywhere.ai/react-native/installation#android-permissions)

Android Permissions

Add required permissions to your `AndroidManifest.xml`:

AndroidManifest.xml

Copy

Ask AI

    <!-- Required for model downloads -->
    <uses-permission android:name="android.permission.INTERNET" />
    
    <!-- Required for STT/VAD with microphone -->
    <uses-permission android:name="android.permission.RECORD_AUDIO" />
    
    <!-- Required for large model storage (optional) -->
    <uses-permission android:name="android.permission.WRITE_EXTERNAL_STORAGE"
        android:maxSdkVersion="28" />
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#running-on-physical-device)

Running on Physical Device

1.  **Enable Developer Options:** Settings → About Phone → Tap “Build Number” 7 times
2.  **Enable USB Debugging:** Settings → Developer Options → USB Debugging
3.  **Connect device:** `adb devices` (should show your device)
4.  **Port forwarding:** `adb reverse tcp:8081 tcp:8081`
5.  **Start Metro:** `npm start`
6.  **Run app:** `npx react-native run-android`

### 

[​](https://docs.runanywhere.ai/react-native/installation#proguard-rules)

ProGuard Rules

If using ProGuard/R8 minification, add these rules to `proguard-rules.pro`:

proguard-rules.pro

Copy

Ask AI

    -keep class com.runanywhere.** { *; }
    -keepclassmembers class com.runanywhere.** { *; }
    -keep class ai.runanywhere.** { *; }
    

[​](https://docs.runanywhere.ai/react-native/installation#troubleshooting-installation)

Troubleshooting Installation
-----------------------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/installation#ios-pod-install-fails)

iOS: Pod Install Fails

Copy

Ask AI

    # Clean and reinstall pods
    cd ios
    rm -rf Pods Podfile.lock
    pod cache clean --all
    pod install
    cd ..
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#android-native-library-not-found)

Android: Native Library Not Found

This error typically occurs when running on an emulator:

Copy

Ask AI

    dlopen failed: library "librunanywherecore.so" not found
    

**Solution:** Use a physical ARM64 Android device. Emulators are not supported. If using a physical device:

Copy

Ask AI

    # Clean and rebuild
    cd android
    ./gradlew clean
    cd ..
    npx react-native run-android
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#android-%E2%80%9Chermesenabled%E2%80%9D-property-not-found)

Android: “hermesEnabled” Property Not Found

**Solution:** Copy the gradle.properties example file:

Copy

Ask AI

    cp android/gradle.properties.example android/gradle.properties
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#android-%E2%80%9Cunable-to-load-script%E2%80%9D-on-device)

Android: “Unable to load script” on Device

**Solution:** Set up port forwarding for Metro bundler:

Copy

Ask AI

    adb reverse tcp:8081 tcp:8081
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#android-grey-screen-after-launch)

Android: Grey Screen After Launch

**Solution:** Metro bundler may have stopped. Restart it:

Copy

Ask AI

    npm start
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#metro-module-resolution-issues)

Metro: Module Resolution Issues

Copy

Ask AI

    # Reset Metro bundler cache
    npx react-native start --reset-cache
    

### 

[​](https://docs.runanywhere.ai/react-native/installation#both-platforms-fresh-build)

Both Platforms: Fresh Build

Copy

Ask AI

    # Complete clean build
    rm -rf node_modules
    npm install
    cd ios && pod install && cd ..
    npx react-native run-ios  # or run-android
    

[​](https://docs.runanywhere.ai/react-native/installation#native-binary-sizes)

Native Binary Sizes
-----------------------------------------------------------------------------------------------------

| Framework | iOS (xcframework) | Android (so) |
| --- | --- | --- |
| RACommons | ~2MB | ~2MB |
| RABackendLLAMACPP | ~15-25MB | ~15-25MB |
| RABackendONNX | ~50-70MB | ~50-70MB |

The ONNX backend is significantly larger due to the ONNX Runtime library. If you only need LLM text generation, consider installing just `@runanywhere/llamacpp` to reduce app size.

[​](https://docs.runanywhere.ai/react-native/installation#next-steps)

Next Steps
-----------------------------------------------------------------------------------

[Quick Start\
-----------\
\
Initialize the SDK and run your first inference →](https://docs.runanywhere.ai/react-native/quick-start)

[Introduction](https://docs.runanywhere.ai/react-native/introduction)
[Quick Start](https://docs.runanywhere.ai/react-native/quick-start)

⌘I

---

# Introduction - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/introduction#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Swift SDK

Introduction

On this page

*   [Overview](https://docs.runanywhere.ai/swift/introduction#overview)
    
*   [Key Capabilities](https://docs.runanywhere.ai/swift/introduction#key-capabilities)
    
*   [Core Philosophy](https://docs.runanywhere.ai/swift/introduction#core-philosophy)
    
*   [Features](https://docs.runanywhere.ai/swift/introduction#features)
    
*   [Language Models (LLM)](https://docs.runanywhere.ai/swift/introduction#language-models-llm)
    
*   [Speech-to-Text (STT)](https://docs.runanywhere.ai/swift/introduction#speech-to-text-stt)
    
*   [Text-to-Speech (TTS)](https://docs.runanywhere.ai/swift/introduction#text-to-speech-tts)
    
*   [Voice Activity Detection (VAD)](https://docs.runanywhere.ai/swift/introduction#voice-activity-detection-vad)
    
*   [Voice Agent Pipeline](https://docs.runanywhere.ai/swift/introduction#voice-agent-pipeline)
    
*   [Model Management](https://docs.runanywhere.ai/swift/introduction#model-management)
    
*   [System Requirements](https://docs.runanywhere.ai/swift/introduction#system-requirements)
    
*   [SDK Modules](https://docs.runanywhere.ai/swift/introduction#sdk-modules)
    
*   [Next Steps](https://docs.runanywhere.ai/swift/introduction#next-steps)
    

[​](https://docs.runanywhere.ai/swift/introduction#overview)

Overview
------------------------------------------------------------------------

The **RunAnywhere Swift SDK** is a production-grade, on-device AI SDK for Apple platforms. It enables developers to run AI models directly on Apple devices without requiring network connectivity for inference, ensuring minimal latency and maximum privacy for your users. The SDK provides a unified interface to multiple AI capabilities:

[LLM\
---\
\
Text generation with streaming support and structured output](https://docs.runanywhere.ai/swift/llm/chat)
[STT\
---\
\
Speech-to-text transcription with multiple backends](https://docs.runanywhere.ai/swift/stt/transcribe)
[TTS\
---\
\
Neural and system voice synthesis](https://docs.runanywhere.ai/swift/tts/synthesize)
[VAD\
---\
\
Real-time voice activity detection](https://docs.runanywhere.ai/swift/vad)

[​](https://docs.runanywhere.ai/swift/introduction#key-capabilities)

Key Capabilities
----------------------------------------------------------------------------------------

*   **Multi-backend architecture** – Choose from LlamaCPP (GGUF models), ONNX Runtime, or Apple Foundation Models
*   **Metal acceleration** – GPU-accelerated inference on Apple Silicon
*   **Event-driven design** – Subscribe to SDK events for reactive UI updates
*   **Production-ready** – Built-in analytics, logging, device registration, and model lifecycle management

[​](https://docs.runanywhere.ai/swift/introduction#core-philosophy)

Core Philosophy
--------------------------------------------------------------------------------------

On-Device First

All AI inference runs locally, ensuring low latency and data privacy. Once models are downloaded, no network connection is required for inference.

Plugin Architecture

Backend engines are optional modules—include only what you need. This keeps your app binary size minimal.

Privacy by Design

Audio and text data never leaves the device unless explicitly configured. Only anonymous analytics are collected by default.

Event-Driven

Subscribe to SDK events for reactive UI updates and observability. Track generation progress, model loading, and errors in real-time.

[​](https://docs.runanywhere.ai/swift/introduction#features)

Features
------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/introduction#language-models-llm)

Language Models (LLM)

*   On-device text generation with streaming support
*   Structured output generation with `Generatable` protocol
*   System prompts and customizable generation parameters
*   Support for thinking/reasoning models with token extraction
*   Multiple framework backends (LlamaCPP, Apple Foundation Models)

### 

[​](https://docs.runanywhere.ai/swift/introduction#speech-to-text-stt)

Speech-to-Text (STT)

*   Real-time streaming transcription
*   Batch audio transcription
*   Multi-language support
*   Whisper-based models via ONNX Runtime

### 

[​](https://docs.runanywhere.ai/swift/introduction#text-to-speech-tts)

Text-to-Speech (TTS)

*   Neural voice synthesis with ONNX models
*   System voices via AVSpeechSynthesizer
*   Streaming audio generation for long text
*   Customizable voice, pitch, rate, and volume

### 

[​](https://docs.runanywhere.ai/swift/introduction#voice-activity-detection-vad)

Voice Activity Detection (VAD)

*   Energy-based speech detection
*   Configurable sensitivity thresholds
*   Real-time audio stream processing

### 

[​](https://docs.runanywhere.ai/swift/introduction#voice-agent-pipeline)

Voice Agent Pipeline

*   Full VAD → STT → LLM → TTS orchestration
*   Complete voice conversation flow
*   Streaming and batch processing modes

### 

[​](https://docs.runanywhere.ai/swift/introduction#model-management)

Model Management

*   Automatic model discovery and catalog sync
*   Download with progress tracking (download, extract, validate stages)
*   In-memory model storage with file system caching
*   Framework-specific model assignment

[​](https://docs.runanywhere.ai/swift/introduction#system-requirements)

System Requirements
----------------------------------------------------------------------------------------------

| Platform | Minimum Version |
| --- | --- |
| iOS | 17.0+ |
| macOS | 14.0+ |
| tvOS | 17.0+ |
| watchOS | 10.0+ |

**Swift Version:** 5.9+ **Xcode:** 15.2+

Some optional modules have higher runtime requirements: - Apple Foundation Models (`RunAnywhereAppleAI`): iOS 26+ / macOS 26+ at runtime

[​](https://docs.runanywhere.ai/swift/introduction#sdk-modules)

SDK Modules
------------------------------------------------------------------------------

| Module | Purpose |
| --- | --- |
| `RunAnywhere` | Core SDK (required) |
| `RunAnywhereLlamaCPP` | LLM text generation with GGUF models |
| `RunAnywhereONNX` | STT/TTS/VAD via ONNX Runtime |

[​](https://docs.runanywhere.ai/swift/introduction#next-steps)

Next Steps
----------------------------------------------------------------------------

[Installation\
------------\
\
Install the SDK via Swift Package Manager](https://docs.runanywhere.ai/swift/installation)
[Quick Start\
-----------\
\
Build your first AI feature in minutes](https://docs.runanywhere.ai/swift/quick-start)

[SDK Overview](https://docs.runanywhere.ai/sdks)
[Installation](https://docs.runanywhere.ai/swift/installation)

⌘I

---

# synthesize() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/tts/synthesize#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

synthesize()

On this page

*   [Basic Synthesis](https://docs.runanywhere.ai/kotlin/tts/synthesize#basic-synthesis)
    
*   [Simple Speak API](https://docs.runanywhere.ai/kotlin/tts/synthesize#simple-speak-api)
    
*   [TTSOutput](https://docs.runanywhere.ai/kotlin/tts/synthesize#ttsoutput)
    
*   [TTSOptions](https://docs.runanywhere.ai/kotlin/tts/synthesize#ttsoptions)
    
*   [Example: Custom Voice Settings](https://docs.runanywhere.ai/kotlin/tts/synthesize#example-custom-voice-settings)
    
*   [Example: Playing Audio with AudioTrack](https://docs.runanywhere.ai/kotlin/tts/synthesize#example-playing-audio-with-audiotrack)
    
*   [Example: Text-to-Speech ViewModel](https://docs.runanywhere.ai/kotlin/tts/synthesize#example-text-to-speech-viewmodel)
    
*   [Check Speaking Status](https://docs.runanywhere.ai/kotlin/tts/synthesize#check-speaking-status)
    
*   [Synthesis Metadata](https://docs.runanywhere.ai/kotlin/tts/synthesize#synthesis-metadata)
    

Synthesize text into audio using on-device neural TTS models.

[​](https://docs.runanywhere.ai/kotlin/tts/synthesize#basic-synthesis)

Basic Synthesis
-----------------------------------------------------------------------------------------

Copy

Ask AI

    // Load a TTS voice first
    RunAnywhere.loadTTSVoice("en-us-default")
    
    // Synthesize text to audio
    val output = RunAnywhere.synthesize(
        text = "Hello, welcome to RunAnywhere!",
        options = TTSOptions(rate = 1.0f, pitch = 1.0f)
    )
    
    // Play the audio
    val audioTrack = createAudioTrack(output.format, output.audioData)
    audioTrack.play()
    

[​](https://docs.runanywhere.ai/kotlin/tts/synthesize#simple-speak-api)

Simple Speak API
-------------------------------------------------------------------------------------------

For convenience, use `speak()` which handles synthesis and playback:

Copy

Ask AI

    // Speak text aloud (handles audio playback automatically)
    val result = RunAnywhere.speak(
        text = "Hello, this is a test message",
        options = TTSOptions(rate = 1.0f)
    )
    
    println("Duration: ${result.duration}s")
    

[​](https://docs.runanywhere.ai/kotlin/tts/synthesize#ttsoutput)

TTSOutput
-----------------------------------------------------------------------------

The synthesis result object:

| Property | Type | Description |
| --- | --- | --- |
| `audioData` | `ByteArray` | Synthesized audio bytes |
| `format` | `AudioFormat` | Audio format (PCM, WAV, etc.) |
| `duration` | `Double` | Duration in seconds |
| `phonemeTimestamps` | `List<TTSPhonemeTimestamp>?` | Phoneme timing |
| `metadata` | `TTSSynthesisMetadata` | Processing metrics |

[​](https://docs.runanywhere.ai/kotlin/tts/synthesize#ttsoptions)

TTSOptions
-------------------------------------------------------------------------------

Customize voice synthesis:

Copy

Ask AI

    data class TTSOptions(
        val voice: String? = null,        // Voice ID (uses loaded voice if null)
        val language: String = "en-US",   // Language code
        val rate: Float = 1.0f,           // Speed (0.0-2.0)
        val pitch: Float = 1.0f,          // Pitch (0.0-2.0)
        val volume: Float = 1.0f,         // Volume (0.0-1.0)
        val audioFormat: AudioFormat = AudioFormat.PCM,
        val sampleRate: Int = 22050,
        val useSSML: Boolean = false      // Enable SSML markup
    )
    

[​](https://docs.runanywhere.ai/kotlin/tts/synthesize#example-custom-voice-settings)

Example: Custom Voice Settings
----------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    val output = RunAnywhere.synthesize(
        text = "The quick brown fox jumps over the lazy dog.",
        options = TTSOptions(
            rate = 0.8f,    // Slower speech
            pitch = 1.2f,   // Higher pitch
            volume = 0.9f
        )
    )
    

[​](https://docs.runanywhere.ai/kotlin/tts/synthesize#example-playing-audio-with-audiotrack)

Example: Playing Audio with AudioTrack
--------------------------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    suspend fun playAudio(audioData: ByteArray, sampleRate: Int = 22050) {
        val audioTrack = AudioTrack.Builder()
            .setAudioAttributes(
                AudioAttributes.Builder()
                    .setUsage(AudioAttributes.USAGE_MEDIA)
                    .setContentType(AudioAttributes.CONTENT_TYPE_SPEECH)
                    .build()
            )
            .setAudioFormat(
                android.media.AudioFormat.Builder()
                    .setEncoding(android.media.AudioFormat.ENCODING_PCM_16BIT)
                    .setSampleRate(sampleRate)
                    .setChannelMask(android.media.AudioFormat.CHANNEL_OUT_MONO)
                    .build()
            )
            .setBufferSizeInBytes(audioData.size)
            .setTransferMode(AudioTrack.MODE_STATIC)
            .build()
    
        audioTrack.write(audioData, 0, audioData.size)
        audioTrack.play()
    }
    

[​](https://docs.runanywhere.ai/kotlin/tts/synthesize#example-text-to-speech-viewmodel)

Example: Text-to-Speech ViewModel
----------------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class TTSViewModel : ViewModel() {
        private val _isSpeaking = MutableStateFlow(false)
        val isSpeaking: StateFlow<Boolean> = _isSpeaking
    
        init {
            viewModelScope.launch {
                // Load TTS voice on init
                RunAnywhere.loadTTSVoice("en-us-default")
            }
        }
    
        fun speak(text: String) {
            viewModelScope.launch {
                _isSpeaking.value = true
                try {
                    RunAnywhere.speak(text, TTSOptions(rate = 1.0f))
                } finally {
                    _isSpeaking.value = false
                }
            }
        }
    
        fun stopSpeaking() {
            viewModelScope.launch {
                RunAnywhere.stopSpeaking()
                _isSpeaking.value = false
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/tts/synthesize#check-speaking-status)

Check Speaking Status
-----------------------------------------------------------------------------------------------------

Copy

Ask AI

    // Check if currently speaking
    val isSpeaking = RunAnywhere.isSpeaking()
    
    // Stop current playback
    RunAnywhere.stopSpeaking()
    

[​](https://docs.runanywhere.ai/kotlin/tts/synthesize#synthesis-metadata)

Synthesis Metadata
-----------------------------------------------------------------------------------------------

Copy

Ask AI

    val output = RunAnywhere.synthesize(text, options)
    
    println("Voice: ${output.metadata.voice}")
    println("Language: ${output.metadata.language}")
    println("Processing time: ${output.metadata.processingTime}s")
    println("Characters: ${output.metadata.characterCount}")
    println("Speed: ${output.metadata.charactersPerSecond} char/s")
    

[STT Options](https://docs.runanywhere.ai/kotlin/stt/options)
[Streaming TTS](https://docs.runanywhere.ai/kotlin/tts/stream)

⌘I

---

# Voice Agent - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/kotlin/voice-agent#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Kotlin SDK

Voice Agent

On this page

*   [Overview](https://docs.runanywhere.ai/kotlin/voice-agent#overview)
    
*   [Configure Voice Agent](https://docs.runanywhere.ai/kotlin/voice-agent#configure-voice-agent)
    
*   [VoiceAgentConfiguration](https://docs.runanywhere.ai/kotlin/voice-agent#voiceagentconfiguration)
    
*   [Process Voice Turn](https://docs.runanywhere.ai/kotlin/voice-agent#process-voice-turn)
    
*   [VoiceAgentResult](https://docs.runanywhere.ai/kotlin/voice-agent#voiceagentresult)
    
*   [Voice Session (Event-Driven)](https://docs.runanywhere.ai/kotlin/voice-agent#voice-session-event-driven)
    
*   [VoiceSessionEvent](https://docs.runanywhere.ai/kotlin/voice-agent#voicesessionevent)
    
*   [Session Control](https://docs.runanywhere.ai/kotlin/voice-agent#session-control)
    
*   [Example: Complete Voice Assistant](https://docs.runanywhere.ai/kotlin/voice-agent#example-complete-voice-assistant)
    
*   [Component States](https://docs.runanywhere.ai/kotlin/voice-agent#component-states)
    
*   [Performance Optimization](https://docs.runanywhere.ai/kotlin/voice-agent#performance-optimization)
    
*   [Typical Latencies (Pixel 7)](https://docs.runanywhere.ai/kotlin/voice-agent#typical-latencies-pixel-7)
    
*   [Related](https://docs.runanywhere.ai/kotlin/voice-agent#related)
    

[​](https://docs.runanywhere.ai/kotlin/voice-agent#overview)

Overview
------------------------------------------------------------------------

The Voice Agent orchestrates the complete voice interaction pipeline: **VAD → STT → LLM → TTS**. This enables full conversational AI with a single API.

[​](https://docs.runanywhere.ai/kotlin/voice-agent#configure-voice-agent)

Configure Voice Agent
--------------------------------------------------------------------------------------------------

Copy

Ask AI

    RunAnywhere.configureVoiceAgent(VoiceAgentConfiguration(
        sttModelId = "whisper-tiny",
        llmModelId = "qwen-0.5b",
        ttsVoiceId = "en-us-default",
        systemPrompt = "You are a helpful assistant. Keep responses brief.",
        interruptionEnabled = true
    ))
    

[​](https://docs.runanywhere.ai/kotlin/voice-agent#voiceagentconfiguration)

VoiceAgentConfiguration
------------------------------------------------------------------------------------------------------

Copy

Ask AI

    data class VoiceAgentConfiguration(
        val sttModelId: String,              // Speech-to-text model
        val llmModelId: String,              // Language model
        val ttsVoiceId: String,              // Text-to-speech voice
        val systemPrompt: String? = null,    // LLM system prompt
        val vadConfiguration: VADConfiguration? = null,
        val interruptionEnabled: Boolean = true  // Allow user interruption
    )
    

[​](https://docs.runanywhere.ai/kotlin/voice-agent#process-voice-turn)

Process Voice Turn
--------------------------------------------------------------------------------------------

Process a complete voice turn (listen → transcribe → respond → speak):

Copy

Ask AI

    val audioData = // ... recorded audio from user
    
    val result = RunAnywhere.processVoice(audioData)
    
    println("User said: ${result.transcription}")
    println("AI response: ${result.response}")
    println("Total latency: ${result.totalLatencyMs}ms")
    
    // Play the response audio
    result.audioData?.let { audio ->
        audioPlayer.play(audio)
    }
    

[​](https://docs.runanywhere.ai/kotlin/voice-agent#voiceagentresult)

VoiceAgentResult
----------------------------------------------------------------------------------------

| Property | Type | Description |
| --- | --- | --- |
| `transcription` | `String` | User’s transcribed speech |
| `response` | `String` | AI’s text response |
| `audioData` | `ByteArray?` | Synthesized response audio |
| `totalLatencyMs` | `Double` | Total processing time |
| `sttLatencyMs` | `Double` | Speech-to-text time |
| `llmLatencyMs` | `Double` | LLM generation time |
| `ttsLatencyMs` | `Double` | Text-to-speech time |

[​](https://docs.runanywhere.ai/kotlin/voice-agent#voice-session-event-driven)

Voice Session (Event-Driven)
--------------------------------------------------------------------------------------------------------------

Start a continuous voice session with event streaming:

Copy

Ask AI

    lifecycleScope.launch {
        RunAnywhere.startVoiceSession()
            .collect { event ->
                when (event) {
                    is VoiceSessionEvent.Listening -> {
                        updateUI("Listening...")
                        showMicrophoneActive()
                    }
                    is VoiceSessionEvent.Transcribed -> {
                        updateUI("You: ${event.text}")
                    }
                    is VoiceSessionEvent.Thinking -> {
                        updateUI("Thinking...")
                        showLoadingIndicator()
                    }
                    is VoiceSessionEvent.Responded -> {
                        updateUI("AI: ${event.text}")
                    }
                    is VoiceSessionEvent.Speaking -> {
                        updateUI("Speaking...")
                    }
                    is VoiceSessionEvent.Idle -> {
                        updateUI("Ready")
                    }
                    is VoiceSessionEvent.Error -> {
                        showError(event.message)
                    }
                }
            }
    }
    

[​](https://docs.runanywhere.ai/kotlin/voice-agent#voicesessionevent)

VoiceSessionEvent
------------------------------------------------------------------------------------------

Copy

Ask AI

    sealed class VoiceSessionEvent {
        object Listening : VoiceSessionEvent()
        data class Transcribed(val text: String) : VoiceSessionEvent()
        object Thinking : VoiceSessionEvent()
        data class Responded(val text: String) : VoiceSessionEvent()
        object Speaking : VoiceSessionEvent()
        object Idle : VoiceSessionEvent()
        data class Error(val message: String) : VoiceSessionEvent()
    }
    

[​](https://docs.runanywhere.ai/kotlin/voice-agent#session-control)

Session Control
--------------------------------------------------------------------------------------

Copy

Ask AI

    // Stop the voice session
    RunAnywhere.stopVoiceSession()
    
    // Check if session is active
    val isActive = RunAnywhere.isVoiceSessionActive()
    
    // Clear conversation history
    RunAnywhere.clearVoiceConversation()
    
    // Update system prompt mid-session
    RunAnywhere.setVoiceSystemPrompt("You are now a coding assistant.")
    

[​](https://docs.runanywhere.ai/kotlin/voice-agent#example-complete-voice-assistant)

Example: Complete Voice Assistant
-------------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class VoiceAssistantActivity : AppCompatActivity() {
    
        override fun onCreate(savedInstanceState: Bundle?) {
            super.onCreate(savedInstanceState)
            setContentView(R.layout.activity_voice_assistant)
    
            lifecycleScope.launch {
                // Configure voice agent
                RunAnywhere.configureVoiceAgent(VoiceAgentConfiguration(
                    sttModelId = "whisper-tiny",
                    llmModelId = "qwen-0.5b",
                    ttsVoiceId = "en-us-default",
                    systemPrompt = """
                        You are a helpful voice assistant.
                        Keep responses under 2 sentences.
                        Be friendly and conversational.
                    """.trimIndent()
                ))
    
                // Check if ready
                if (!RunAnywhere.isVoiceAgentReady()) {
                    showMessage("Loading models...")
                    RunAnywhere.initializeVoiceAgentWithLoadedModels()
                }
            }
    
            startButton.setOnClickListener {
                startVoiceSession()
            }
    
            stopButton.setOnClickListener {
                lifecycleScope.launch {
                    RunAnywhere.stopVoiceSession()
                }
            }
        }
    
        private fun startVoiceSession() {
            lifecycleScope.launch {
                RunAnywhere.startVoiceSession()
                    .collect { event ->
                        runOnUiThread {
                            when (event) {
                                is VoiceSessionEvent.Listening -> {
                                    statusText.text = "🎤 Listening..."
                                    micButton.setColorFilter(Color.RED)
                                }
                                is VoiceSessionEvent.Transcribed -> {
                                    transcriptionText.text = event.text
                                }
                                is VoiceSessionEvent.Thinking -> {
                                    statusText.text = "🤔 Thinking..."
                                }
                                is VoiceSessionEvent.Responded -> {
                                    responseText.text = event.text
                                }
                                is VoiceSessionEvent.Speaking -> {
                                    statusText.text = "🔊 Speaking..."
                                }
                                is VoiceSessionEvent.Idle -> {
                                    statusText.text = "Ready"
                                    micButton.clearColorFilter()
                                }
                                is VoiceSessionEvent.Error -> {
                                    Toast.makeText(this@VoiceAssistantActivity,
                                        event.message, Toast.LENGTH_SHORT).show()
                                }
                            }
                        }
                    }
            }
        }
    }
    

[​](https://docs.runanywhere.ai/kotlin/voice-agent#component-states)

Component States
----------------------------------------------------------------------------------------

Check individual component readiness:

Copy

Ask AI

    val states = RunAnywhere.voiceAgentComponentStates()
    println("STT ready: ${states.sttReady}")
    println("LLM ready: ${states.llmReady}")
    println("TTS ready: ${states.ttsReady}")
    println("VAD ready: ${states.vadReady}")
    

[​](https://docs.runanywhere.ai/kotlin/voice-agent#performance-optimization)

Performance Optimization
--------------------------------------------------------------------------------------------------------

**Minimize total latency:** Use smaller models (whisper-tiny, 0.5B LLM), enable streaming for both LLM and TTS, and preload all models at app startup.

### 

[​](https://docs.runanywhere.ai/kotlin/voice-agent#typical-latencies-pixel-7)

Typical Latencies (Pixel 7)

| Stage | Latency |
| --- | --- |
| STT | 200-400ms for 5s audio |
| LLM | 150-300ms for 50 tokens |
| TTS | 100-200ms for short responses |
| **Total round-trip** | **500-900ms** |

[​](https://docs.runanywhere.ai/kotlin/voice-agent#related)

Related
----------------------------------------------------------------------

[VAD\
---\
\
Voice Activity Detection](https://docs.runanywhere.ai/kotlin/vad)
[STT\
---\
\
Speech-to-Text](https://docs.runanywhere.ai/kotlin/stt/transcribe)
[TTS\
---\
\
Text-to-Speech](https://docs.runanywhere.ai/kotlin/tts/synthesize)
[LLM\
---\
\
Text Generation](https://docs.runanywhere.ai/kotlin/llm/chat)

[Voice Activity Detection](https://docs.runanywhere.ai/kotlin/vad)
[Configuration](https://docs.runanywhere.ai/kotlin/configuration)

⌘I

---

# chat() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/llm/chat#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

chat()

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/llm/chat#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/react-native/llm/chat#basic-usage)
    
*   [API Reference](https://docs.runanywhere.ai/react-native/llm/chat#api-reference)
    
*   [Parameters](https://docs.runanywhere.ai/react-native/llm/chat#parameters)
    
*   [Returns](https://docs.runanywhere.ai/react-native/llm/chat#returns)
    
*   [Throws](https://docs.runanywhere.ai/react-native/llm/chat#throws)
    
*   [Examples](https://docs.runanywhere.ai/react-native/llm/chat#examples)
    
*   [Simple Q&A](https://docs.runanywhere.ai/react-native/llm/chat#simple-q%26a)
    
*   [React Component](https://docs.runanywhere.ai/react-native/llm/chat#react-component)
    
*   [With Error Handling](https://docs.runanywhere.ai/react-native/llm/chat#with-error-handling)
    
*   [Chat vs Generate](https://docs.runanywhere.ai/react-native/llm/chat#chat-vs-generate)
    
*   [Related](https://docs.runanywhere.ai/react-native/llm/chat#related)
    

[​](https://docs.runanywhere.ai/react-native/llm/chat#overview)

Overview
---------------------------------------------------------------------------

The `chat()` method provides the simplest way to interact with an LLM. It takes a prompt and returns just the response text, making it perfect for quick integrations.

[​](https://docs.runanywhere.ai/react-native/llm/chat#basic-usage)

Basic Usage
---------------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    // Simple question
    const response = await RunAnywhere.chat('What is the capital of France?')
    console.log(response) // "Paris is the capital of France."
    
    // Follow-up (stateless - each call is independent)
    const answer = await RunAnywhere.chat('What is 2 + 2?')
    console.log(answer) // "4"
    

[​](https://docs.runanywhere.ai/react-native/llm/chat#api-reference)

API Reference
-------------------------------------------------------------------------------------

Copy

Ask AI

    await RunAnywhere.chat(prompt: string): Promise<string>
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/chat#parameters)

Parameters

| Parameter | Type | Description |
| --- | --- | --- |
| `prompt` | `string` | The user’s message or question |

### 

[​](https://docs.runanywhere.ai/react-native/llm/chat#returns)

Returns

| Type | Description |
| --- | --- |
| `Promise<string>` | The generated response text |

### 

[​](https://docs.runanywhere.ai/react-native/llm/chat#throws)

Throws

| Error Code | Description |
| --- | --- |
| `notInitialized` | SDK not initialized |
| `modelNotLoaded` | No LLM model loaded |
| `generationFailed` | Generation failed |

[​](https://docs.runanywhere.ai/react-native/llm/chat#examples)

Examples
---------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/llm/chat#simple-q&a)

Simple Q&A

Copy

Ask AI

    const capital = await RunAnywhere.chat('What is the capital of Japan?')
    // "Tokyo is the capital of Japan."
    
    const math = await RunAnywhere.chat('Calculate 15% of 200')
    // "15% of 200 is 30."
    
    const translation = await RunAnywhere.chat('Translate "Hello" to Spanish')
    // "Hola"
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/chat#react-component)

React Component

ChatComponent.tsx

Copy

Ask AI

    import React, { useState } from 'react'
    import { View, TextInput, Button, Text } from 'react-native'
    import { RunAnywhere } from '@runanywhere/core'
    
    export function ChatComponent() {
      const [input, setInput] = useState('')
      const [response, setResponse] = useState('')
      const [loading, setLoading] = useState(false)
    
      const handleSend = async () => {
        if (!input.trim()) return
    
        setLoading(true)
        try {
          const answer = await RunAnywhere.chat(input)
          setResponse(answer)
        } catch (error) {
          setResponse('Error: ' + (error as Error).message)
        } finally {
          setLoading(false)
        }
      }
    
      return (
        <View style={{ padding: 16 }}>
          <TextInput
            value={input}
            onChangeText={setInput}
            placeholder="Ask anything..."
            style={{ borderWidth: 1, padding: 8, marginBottom: 8 }}
          />
          <Button title={loading ? 'Thinking...' : 'Send'} onPress={handleSend} disabled={loading} />
          {response ? <Text style={{ marginTop: 16 }}>{response}</Text> : null}
        </View>
      )
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/chat#with-error-handling)

With Error Handling

Copy

Ask AI

    import { RunAnywhere, isSDKError, SDKErrorCode } from '@runanywhere/core'
    
    async function askQuestion(question: string): Promise<string> {
      try {
        return await RunAnywhere.chat(question)
      } catch (error) {
        if (isSDKError(error)) {
          switch (error.code) {
            case SDKErrorCode.notInitialized:
              throw new Error('Please initialize the SDK first')
            case SDKErrorCode.modelNotLoaded:
              throw new Error('Please load a model first')
            default:
              throw new Error('AI error: ' + error.message)
          }
        }
        throw error
      }
    }
    
    // Usage
    const response = await askQuestion('What is AI?')
    

[​](https://docs.runanywhere.ai/react-native/llm/chat#chat-vs-generate)

Chat vs Generate
-------------------------------------------------------------------------------------------

| Feature | `chat()` | `generate()` |
| --- | --- | --- |
| Return type | `string` | `GenerationResult` |
| Metrics | No  | Yes (tokens, latency, tok/s) |
| Options | No  | Yes (maxTokens, temperature, etc.) |
| System prompt | No  | Yes |
| Use case | Quick prototyping | Production apps |

Use `chat()` for quick prototyping and simple interactions. Switch to `generate()` when you need performance metrics, custom options, or system prompts.

[​](https://docs.runanywhere.ai/react-native/llm/chat#related)

Related
-------------------------------------------------------------------------

[Generate\
--------\
\
Full generation with options and metrics](https://docs.runanywhere.ai/react-native/llm/generate)
[Streaming\
---------\
\
Real-time token streaming](https://docs.runanywhere.ai/react-native/llm/stream)

[Quick Start](https://docs.runanywhere.ai/react-native/quick-start)
[generate()](https://docs.runanywhere.ai/react-native/llm/generate)

⌘I

---

# Quick Start - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/quickstart#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Getting Started

Quick Start

On this page

*   [Choose Your Platform](https://docs.runanywhere.ai/quickstart#choose-your-platform)
    
*   [Next Steps](https://docs.runanywhere.ai/quickstart#next-steps)
    

[​](https://docs.runanywhere.ai/quickstart#choose-your-platform)

Choose Your Platform
----------------------------------------------------------------------------------------

Select your development platform to get started:

*   Swift
    
*   Kotlin
    
*   React Native
    
*   Flutter
    

### 

[​](https://docs.runanywhere.ai/quickstart#1-install-the-sdk)

1\. Install the SDK

Add RunAnywhere to your project using Swift Package Manager:

Package.swift

Copy

Ask AI

    dependencies: [\
        .package(url: "https://github.com/RunanywhereAI/runanywhere-sdks.git", from: "0.1.0")\
    ]
    

Add the products you need:

Copy

Ask AI

    .target(
        name: "YourApp",
        dependencies: [\
            .product(name: "RunAnywhere", package: "runanywhere-swift"),\
            .product(name: "RunAnywhereLlamaCPP", package: "runanywhere-swift"),\
            .product(name: "RunAnywhereONNX", package: "runanywhere-swift"),\
        ]
    )
    

### 

[​](https://docs.runanywhere.ai/quickstart#2-initialize-the-sdk)

2\. Initialize the SDK

Copy

Ask AI

    import RunAnywhere
    import LlamaCPPRuntime
    
    @main
    struct MyApp: App {
        init() {
            do {
                try RunAnywhere.initialize(
                    apiKey: "your-api-key",
                    baseURL: "https://api.runanywhere.ai",
                    environment: .production
                )
    
                Task { @MainActor in
                    ModuleRegistry.shared.register(LlamaCPP.self)
                }
            } catch {
                print("SDK init failed: \(error)")
            }
        }
    
        var body: some Scene {
            WindowGroup {
                ContentView()
            }
        }
    }
    

### 

[​](https://docs.runanywhere.ai/quickstart#3-generate-text)

3\. Generate Text

Copy

Ask AI

    let response = try await llm.chat(
        prompt: "Hello, how can you help me today?",
        options: .init(maxTokens: 256, temperature: 0.7)
    )
    print(response.text)
    

[Full Swift Documentation\
------------------------\
\
Explore all Swift SDK features →](https://docs.runanywhere.ai/swift/introduction)

### 

[​](https://docs.runanywhere.ai/quickstart#1-install-the-sdk-2)

1\. Install the SDK

Add RunAnywhere to your `build.gradle.kts`:

Copy

Ask AI

    dependencies {
        implementation("ai.runanywhere:runanywhere-core:0.1.0")
        implementation("ai.runanywhere:runanywhere-llamacpp:0.1.0")
        implementation("ai.runanywhere:runanywhere-onnx:0.1.0")
    }
    

### 

[​](https://docs.runanywhere.ai/quickstart#2-initialize-the-sdk-2)

2\. Initialize the SDK

Copy

Ask AI

    import ai.runanywhere.RunAnywhere
    import ai.runanywhere.llamacpp.LlamaCPP
    
    class MyApplication : Application() {
        override fun onCreate() {
            super.onCreate()
    
            RunAnywhere.initialize(
                apiKey = "your-api-key",
                baseURL = "https://api.runanywhere.ai",
                environment = Environment.PRODUCTION
            )
    
            ModuleRegistry.register(LlamaCPP::class)
        }
    }
    

### 

[​](https://docs.runanywhere.ai/quickstart#3-generate-text-2)

3\. Generate Text

Copy

Ask AI

    val response = llm.chat(
        prompt = "Hello, how can you help me today?",
        options = ChatOptions(maxTokens = 256, temperature = 0.7f)
    )
    println(response.text)
    

[Full Kotlin Documentation\
-------------------------\
\
Explore all Kotlin SDK features →](https://docs.runanywhere.ai/kotlin/introduction)

### 

[​](https://docs.runanywhere.ai/quickstart#1-install-the-sdk-3)

1\. Install the SDK

Copy

Ask AI

    npm install @runanywhere/react-native
    # or
    yarn add @runanywhere/react-native
    

### 

[​](https://docs.runanywhere.ai/quickstart#2-initialize-the-sdk-3)

2\. Initialize the SDK

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/react-native';
    
    // In your app entry point
    RunAnywhere.initialize({
      apiKey: 'your-api-key',
      baseURL: 'https://api.runanywhere.ai',
      environment: 'production',
    });
    

### 

[​](https://docs.runanywhere.ai/quickstart#3-generate-text-3)

3\. Generate Text

Copy

Ask AI

    import { useLLM } from '@runanywhere/react-native';
    
    function MyComponent() {
      const { chat } = useLLM();
    
      const handleChat = async () => {
        const response = await chat({
          prompt: 'Hello, how can you help me today?',
          options: { maxTokens: 256, temperature: 0.7 },
        });
        console.log(response.text);
      };
    
      return <Button onPress={handleChat} title="Chat" />;
    }
    

[Full React Native Documentation\
-------------------------------\
\
Explore all React Native SDK features →](https://docs.runanywhere.ai/react-native/introduction)

### 

[​](https://docs.runanywhere.ai/quickstart#1-install-the-sdk-4)

1\. Install the SDK

Add to your `pubspec.yaml`:

Copy

Ask AI

    dependencies:
      runanywhere: ^0.1.0
    

Then run:

Copy

Ask AI

    flutter pub get
    

### 

[​](https://docs.runanywhere.ai/quickstart#2-initialize-the-sdk-4)

2\. Initialize the SDK

Copy

Ask AI

    import 'package:runanywhere/runanywhere.dart';
    
    void main() async {
      WidgetsFlutterBinding.ensureInitialized();
    
      await RunAnywhere.initialize(
        apiKey: 'your-api-key',
        baseURL: 'https://api.runanywhere.ai',
        environment: Environment.production,
      );
    
      runApp(MyApp());
    }
    

### 

[​](https://docs.runanywhere.ai/quickstart#3-generate-text-4)

3\. Generate Text

Copy

Ask AI

    final response = await llm.chat(
      prompt: 'Hello, how can you help me today?',
      options: ChatOptions(maxTokens: 256, temperature: 0.7),
    );
    print(response.text);
    

[Full Flutter Documentation\
--------------------------\
\
Explore all Flutter SDK features →](https://docs.runanywhere.ai/flutter/introduction)

* * *

[​](https://docs.runanywhere.ai/quickstart#next-steps)

Next Steps
--------------------------------------------------------------------

[LLM Guide\
---------\
\
Learn about text generation and chat](https://docs.runanywhere.ai/swift/llm/chat)
[Speech-to-Text\
--------------\
\
Add voice transcription to your app](https://docs.runanywhere.ai/swift/stt/transcribe)
[Text-to-Speech\
--------------\
\
Give your app a voice](https://docs.runanywhere.ai/swift/tts/synthesize)
[Voice Agent\
-----------\
\
Build complete voice experiences](https://docs.runanywhere.ai/swift/voice-agent)

**Need help?** Check out the [SDK Overview](https://docs.runanywhere.ai/sdks)
 to compare all platforms, or dive into the full documentation for your chosen SDK.

[Welcome to RunAnywhere](https://docs.runanywhere.ai/)
[SDK Overview](https://docs.runanywhere.ai/sdks)

⌘I

---

# Voice Agent - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/flutter/voice-agent#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Flutter SDK

Voice Agent

On this page

*   [Overview](https://docs.runanywhere.ai/flutter/voice-agent#overview)
    
*   [Prerequisites](https://docs.runanywhere.ai/flutter/voice-agent#prerequisites)
    
*   [Basic Usage](https://docs.runanywhere.ai/flutter/voice-agent#basic-usage)
    
*   [VoiceSessionConfig](https://docs.runanywhere.ai/flutter/voice-agent#voicesessionconfig)
    
*   [VoiceSessionHandle](https://docs.runanywhere.ai/flutter/voice-agent#voicesessionhandle)
    
*   [Voice Session Events](https://docs.runanywhere.ai/flutter/voice-agent#voice-session-events)
    
*   [Complete Example](https://docs.runanywhere.ai/flutter/voice-agent#complete-example)
    
*   [Push-to-Talk Mode](https://docs.runanywhere.ai/flutter/voice-agent#push-to-talk-mode)
    
*   [Component States](https://docs.runanywhere.ai/flutter/voice-agent#component-states)
    
*   [Cleanup](https://docs.runanywhere.ai/flutter/voice-agent#cleanup)
    
*   [See Also](https://docs.runanywhere.ai/flutter/voice-agent#see-also)
    

The Voice Agent orchestrates the complete voice interaction pipeline: **VAD → STT → LLM → TTS**. Build hands-free voice assistants with a single API.

See the Voice Agent in action! The [Flutter Starter Example](https://github.com/RunanywhereAI/flutter-starter-example)
 includes a complete voice assistant implementation.

[​](https://docs.runanywhere.ai/flutter/voice-agent#overview)

Overview
-------------------------------------------------------------------------

The pipeline automatically:

1.  **Listens** for speech using VAD
2.  **Transcribes** what the user says
3.  **Generates** an AI response
4.  **Speaks** the response back

[​](https://docs.runanywhere.ai/flutter/voice-agent#prerequisites)

Prerequisites
-----------------------------------------------------------------------------------

Load all required models before starting a voice session:

Copy

Ask AI

    // Check if ready
    if (!RunAnywhere.isVoiceAgentReady) {
      await RunAnywhere.loadSTTModel('sherpa-onnx-whisper-tiny.en');
      await RunAnywhere.loadModel('smollm2-360m-q8_0');
      await RunAnywhere.loadTTSVoice('vits-piper-en_US-lessac-medium');
    }
    

[​](https://docs.runanywhere.ai/flutter/voice-agent#basic-usage)

Basic Usage
-------------------------------------------------------------------------------

Copy

Ask AI

    // Start voice session
    final session = await RunAnywhere.startVoiceSession();
    
    // Listen to events
    session.events.listen((event) {
      switch (event) {
        case VoiceSessionListening(:final audioLevel):
          print('Listening... Level: $audioLevel');
        case VoiceSessionSpeechStarted():
          print('Speech detected!');
        case VoiceSessionTranscribed(:final text):
          print('User: $text');
        case VoiceSessionResponded(:final text):
          print('AI: $text');
        case VoiceSessionTurnCompleted(:final transcript, :final response):
          print('Turn complete - User: $transcript, AI: $response');
      }
    });
    
    // Stop when done
    session.stop();
    

[​](https://docs.runanywhere.ai/flutter/voice-agent#voicesessionconfig)

VoiceSessionConfig
---------------------------------------------------------------------------------------------

Configure voice session behavior:

Copy

Ask AI

    final session = await RunAnywhere.startVoiceSession(
      config: VoiceSessionConfig(
        silenceDuration: 1.5,     // Seconds of silence before processing
        speechThreshold: 0.03,    // Audio level threshold for speech
        autoPlayTTS: true,        // Automatically play TTS response
        continuousMode: true,     // Resume listening after TTS completes
      ),
    );
    

| Parameter | Type | Default | Description |
| --- | --- | --- | --- |
| `silenceDuration` | `double` | 1.5 | Seconds of silence to trigger processing |
| `speechThreshold` | `double` | 0.03 | Audio level threshold for speech detection |
| `autoPlayTTS` | `bool` | true | Auto-play synthesized audio |
| `continuousMode` | `bool` | true | Keep listening after response |

[​](https://docs.runanywhere.ai/flutter/voice-agent#voicesessionhandle)

VoiceSessionHandle
---------------------------------------------------------------------------------------------

Control the voice session:

Copy

Ask AI

    final session = await RunAnywhere.startVoiceSession();
    
    // Check state
    print('Running: ${session.isRunning}');
    print('Processing: ${session.isProcessing}');
    
    // Start listening (called automatically)
    await session.start();
    
    // Force process current audio (push-to-talk mode)
    await session.sendNow();
    
    // Stop session
    session.stop();
    
    // Clean up
    await session.dispose();
    

[​](https://docs.runanywhere.ai/flutter/voice-agent#voice-session-events)

Voice Session Events
-------------------------------------------------------------------------------------------------

| Event | Description | Properties |
| --- | --- | --- |
| `VoiceSessionStarted` | Session started | \-  |
| `VoiceSessionListening` | Actively listening | `audioLevel` |
| `VoiceSessionSpeechStarted` | Speech detected | \-  |
| `VoiceSessionProcessing` | Processing audio | \-  |
| `VoiceSessionTranscribed` | Got transcription | `text` |
| `VoiceSessionResponded` | Got LLM response | `text` |
| `VoiceSessionSpeaking` | Playing TTS | \-  |
| `VoiceSessionTurnCompleted` | Turn finished | `transcript`, `response` |
| `VoiceSessionStopped` | Session stopped | \-  |
| `VoiceSessionError` | Error occurred | `message` |

[​](https://docs.runanywhere.ai/flutter/voice-agent#complete-example)

Complete Example
-----------------------------------------------------------------------------------------

Copy

Ask AI

    class VoiceAssistantScreen extends StatefulWidget {
      @override
      _VoiceAssistantScreenState createState() => _VoiceAssistantScreenState();
    }
    
    class _VoiceAssistantScreenState extends State<VoiceAssistantScreen> {
      VoiceSessionHandle? _session;
      List<ChatMessage> _messages = [];
      double _audioLevel = 0.0;
      String _status = 'Tap to start';
      bool _isActive = false;
    
      Future<void> _initializeModels() async {
        setState(() => _status = 'Loading models...');
    
        // Download and load all required models
        if (!RunAnywhere.isSTTModelLoaded) {
          await _downloadAndLoad('sherpa-onnx-whisper-tiny.en', 'STT');
          await RunAnywhere.loadSTTModel('sherpa-onnx-whisper-tiny.en');
        }
    
        if (!RunAnywhere.isModelLoaded) {
          await _downloadAndLoad('smollm2-360m-q8_0', 'LLM');
          await RunAnywhere.loadModel('smollm2-360m-q8_0');
        }
    
        if (!RunAnywhere.isTTSVoiceLoaded) {
          await _downloadAndLoad('vits-piper-en_US-lessac-medium', 'TTS');
          await RunAnywhere.loadTTSVoice('vits-piper-en_US-lessac-medium');
        }
    
        setState(() => _status = 'Ready');
      }
    
      Future<void> _downloadAndLoad(String modelId, String label) async {
        await for (final p in RunAnywhere.downloadModel(modelId)) {
          setState(() => _status = 'Downloading $label: ${(p.percentage * 100).toStringAsFixed(0)}%');
          if (p.state.isCompleted) break;
        }
      }
    
      Future<void> _toggleSession() async {
        if (_session != null) {
          _session!.stop();
          setState(() {
            _session = null;
            _isActive = false;
            _status = 'Stopped';
          });
          return;
        }
    
        // Ensure models are loaded
        if (!RunAnywhere.isVoiceAgentReady) {
          await _initializeModels();
        }
    
        // Start session
        _session = await RunAnywhere.startVoiceSession(
          config: VoiceSessionConfig(
            silenceDuration: 1.5,
            autoPlayTTS: true,
            continuousMode: true,
          ),
        );
    
        setState(() => _isActive = true);
    
        // Handle events
        _session!.events.listen((event) {
          setState(() {
            switch (event) {
              case VoiceSessionListening(:final audioLevel):
                _audioLevel = audioLevel;
                _status = 'Listening...';
    
              case VoiceSessionSpeechStarted():
                _status = 'Speech detected';
    
              case VoiceSessionProcessing():
                _status = 'Processing...';
    
              case VoiceSessionTranscribed(:final text):
                _messages.add(ChatMessage(text: text, isUser: true));
    
              case VoiceSessionResponded(:final text):
                _messages.add(ChatMessage(text: text, isUser: false));
    
              case VoiceSessionSpeaking():
                _status = 'Speaking...';
    
              case VoiceSessionTurnCompleted():
                _status = 'Listening...';
    
              case VoiceSessionError(:final message):
                _status = 'Error: $message';
    
              case VoiceSessionStopped():
                _status = 'Stopped';
                _isActive = false;
            }
          });
        });
      }
    
      @override
      Widget build(BuildContext context) {
        return Scaffold(
          appBar: AppBar(title: Text('Voice Assistant')),
          body: Column(
            children: [\
              // Chat messages\
              Expanded(\
                child: ListView.builder(\
                  itemCount: _messages.length,\
                  itemBuilder: (context, index) {\
                    final msg = _messages[index];\
                    return Align(\
                      alignment: msg.isUser\
                          ? Alignment.centerRight\
                          : Alignment.centerLeft,\
                      child: Container(\
                        margin: EdgeInsets.all(8),\
                        padding: EdgeInsets.all(12),\
                        decoration: BoxDecoration(\
                          color: msg.isUser ? Colors.blue : Colors.grey[300],\
                          borderRadius: BorderRadius.circular(12),\
                        ),\
                        child: Text(\
                          msg.text,\
                          style: TextStyle(\
                            color: msg.isUser ? Colors.white : Colors.black,\
                          ),\
                        ),\
                      ),\
                    );\
                  },\
                ),\
              ),\
    \
              // Status and controls\
              Container(\
                padding: EdgeInsets.all(16),\
                child: Column(\
                  children: [\
                    // Audio level indicator\
                    if (_isActive)\
                      Container(\
                        height: 4,\
                        child: LinearProgressIndicator(\
                          value: _audioLevel.clamp(0.0, 1.0),\
                        ),\
                      ),\
    \
                    SizedBox(height: 8),\
                    Text(_status),\
                    SizedBox(height: 16),\
    \
                    // Start/Stop button\
                    FloatingActionButton(\
                      onPressed: _toggleSession,\
                      backgroundColor: _isActive ? Colors.red : Colors.blue,\
                      child: Icon(_isActive ? Icons.stop : Icons.mic),\
                    ),\
                  ],\
                ),\
              ),\
            ],
          ),
        );
      }
    
      @override
      void dispose() {
        _session?.stop();
        super.dispose();
      }
    }
    
    class ChatMessage {
      final String text;
      final bool isUser;
    
      ChatMessage({required this.text, required this.isUser});
    }
    

[​](https://docs.runanywhere.ai/flutter/voice-agent#push-to-talk-mode)

Push-to-Talk Mode
-------------------------------------------------------------------------------------------

For push-to-talk instead of continuous listening:

Copy

Ask AI

    final session = await RunAnywhere.startVoiceSession(
      config: VoiceSessionConfig(
        continuousMode: false,  // Don't auto-resume
      ),
    );
    
    // Manual control
    void onButtonDown() async {
      await session.start();
    }
    
    void onButtonUp() async {
      await session.sendNow();  // Force process current audio
    }
    

[​](https://docs.runanywhere.ai/flutter/voice-agent#component-states)

Component States
-----------------------------------------------------------------------------------------

Check individual component states:

Copy

Ask AI

    final states = RunAnywhere.getVoiceAgentComponentStates();
    
    print('STT: ${states.stt}');    // loaded, loading, unloaded, error
    print('LLM: ${states.llm}');
    print('TTS: ${states.tts}');
    print('All ready: ${states.isFullyReady}');
    

[​](https://docs.runanywhere.ai/flutter/voice-agent#cleanup)

Cleanup
-----------------------------------------------------------------------

Copy

Ask AI

    // Stop session
    session.stop();
    
    // Cleanup resources
    RunAnywhere.cleanupVoiceAgent();
    
    // Optionally unload models to free memory
    await RunAnywhere.unloadSTTModel();
    await RunAnywhere.unloadModel();
    await RunAnywhere.unloadTTSVoice();
    

[​](https://docs.runanywhere.ai/flutter/voice-agent#see-also)

See Also
-------------------------------------------------------------------------

[VAD\
---\
\
Voice activity detection](https://docs.runanywhere.ai/flutter/vad)
[STT\
---\
\
Speech-to-text](https://docs.runanywhere.ai/flutter/stt/transcribe)
[LLM\
---\
\
Text generation](https://docs.runanywhere.ai/flutter/llm/chat)
[TTS\
---\
\
Text-to-speech](https://docs.runanywhere.ai/flutter/tts/synthesize)

[Voice Activity Detection](https://docs.runanywhere.ai/flutter/vad)
[Configuration](https://docs.runanywhere.ai/flutter/configuration)

⌘I

---

# Configuration - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/configuration#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Swift SDK

Configuration

On this page

*   [SDK Initialization](https://docs.runanywhere.ai/swift/configuration#sdk-initialization)
    
*   [Parameters](https://docs.runanywhere.ai/swift/configuration#parameters)
    
*   [SDKEnvironment](https://docs.runanywhere.ai/swift/configuration#sdkenvironment)
    
*   [Module Registration](https://docs.runanywhere.ai/swift/configuration#module-registration)
    
*   [Priority](https://docs.runanywhere.ai/swift/configuration#priority)
    
*   [LLM Generation Options](https://docs.runanywhere.ai/swift/configuration#llm-generation-options)
    
*   [Default Values](https://docs.runanywhere.ai/swift/configuration#default-values)
    
*   [Example Configurations](https://docs.runanywhere.ai/swift/configuration#example-configurations)
    
*   [STT Options](https://docs.runanywhere.ai/swift/configuration#stt-options)
    
*   [TTS Options](https://docs.runanywhere.ai/swift/configuration#tts-options)
    
*   [VAD Configuration](https://docs.runanywhere.ai/swift/configuration#vad-configuration)
    
*   [Memory Considerations](https://docs.runanywhere.ai/swift/configuration#memory-considerations)
    
*   [Model Memory Usage](https://docs.runanywhere.ai/swift/configuration#model-memory-usage)
    
*   [Memory Management](https://docs.runanywhere.ai/swift/configuration#memory-management)
    
*   [Logging Configuration](https://docs.runanywhere.ai/swift/configuration#logging-configuration)
    
*   [Log Levels](https://docs.runanywhere.ai/swift/configuration#log-levels)
    
*   [SDK State](https://docs.runanywhere.ai/swift/configuration#sdk-state)
    
*   [Environment-Specific Settings](https://docs.runanywhere.ai/swift/configuration#environment-specific-settings)
    
*   [Reset (Testing)](https://docs.runanywhere.ai/swift/configuration#reset-testing)
    

This guide covers all configuration options for the RunAnywhere Swift SDK.

[​](https://docs.runanywhere.ai/swift/configuration#sdk-initialization)

SDK Initialization
---------------------------------------------------------------------------------------------

Copy

Ask AI

    try RunAnywhere.initialize(
        apiKey: "<YOUR_API_KEY>",
        baseURL: "https://api.runanywhere.ai",
        environment: .production
    )
    

### 

[​](https://docs.runanywhere.ai/swift/configuration#parameters)

Parameters

| Parameter | Type | Required | Description |
| --- | --- | --- | --- |
| `apiKey` | `String?` | Production | API key for authentication |
| `baseURL` | `String?` | Production | Backend API base URL |
| `environment` | `SDKEnvironment` | No  | Environment mode (default: .development) |

[​](https://docs.runanywhere.ai/swift/configuration#sdkenvironment)

SDKEnvironment
-------------------------------------------------------------------------------------

Copy

Ask AI

    public enum SDKEnvironment: String, Sendable {
        case development
        case staging
        case production
    }
    

| Environment | Log Level | Authentication | Analytics | Description |
| --- | --- | --- | --- | --- |
| `.development` | Debug | Optional | Local | Verbose logging, mock services |
| `.staging` | Info | Required | Yes | Testing with real services |
| `.production` | Warning | Required | Yes | Minimal logging, full auth |

[​](https://docs.runanywhere.ai/swift/configuration#module-registration)

Module Registration
-----------------------------------------------------------------------------------------------

Register modules at app startup to enable capabilities:

Copy

Ask AI

    import RunAnywhere
    import LlamaCPPRuntime
    import ONNXRuntime
    
    @MainActor
    func configureSDK() {
        // Register LlamaCPP for LLM capability
        LlamaCPP.register(priority: 100)
    
        // Register ONNX for STT/TTS/VAD
        ONNX.register(priority: 100)
    }
    

### 

[​](https://docs.runanywhere.ai/swift/configuration#priority)

Priority

Higher priority modules are preferred when multiple can handle a request:

Copy

Ask AI

    // Custom module with higher priority
    MyCustomLLM.register(priority: 200)  // Used first
    LlamaCPP.register(priority: 100)     // Fallback
    

[​](https://docs.runanywhere.ai/swift/configuration#llm-generation-options)

LLM Generation Options
-----------------------------------------------------------------------------------------------------

Copy

Ask AI

    public struct LLMGenerationOptions: Sendable {
        public let maxTokens: Int               // Maximum tokens to generate
        public let temperature: Float           // Randomness (0.0-2.0)
        public let topP: Float                  // Nucleus sampling (0.0-1.0)
        public let stopSequences: [String]      // Stop generation strings
        public let streamingEnabled: Bool       // Token streaming
        public let preferredFramework: InferenceFramework?
        public let structuredOutput: StructuredOutputConfig?
        public let systemPrompt: String?        // System prompt
    }
    

### 

[​](https://docs.runanywhere.ai/swift/configuration#default-values)

Default Values

| Option | Default | Range/Type | Description |
| --- | --- | --- | --- |
| `maxTokens` | 100 | 1-4096+ | Max tokens to generate |
| `temperature` | 0.8 | 0.0-2.0 | Lower = deterministic |
| `topP` | 1.0 | 0.0-1.0 | Nucleus sampling threshold |
| `stopSequences` | `[]` | `[String]` | Stop on these strings |
| `streamingEnabled` | false | `Bool` | Enable token streaming |
| `systemPrompt` | nil | `String?` | System instruction |

### 

[​](https://docs.runanywhere.ai/swift/configuration#example-configurations)

Example Configurations

Copy

Ask AI

    // Deterministic, factual responses
    let factual = LLMGenerationOptions(
        maxTokens: 200,
        temperature: 0.1,
        topP: 0.9
    )
    
    // Creative writing
    let creative = LLMGenerationOptions(
        maxTokens: 500,
        temperature: 1.2,
        topP: 0.95
    )
    
    // JSON output
    let json = LLMGenerationOptions(
        maxTokens: 300,
        temperature: 0.2,
        stopSequences: ["}"],
        systemPrompt: "Respond only with valid JSON."
    )
    

[​](https://docs.runanywhere.ai/swift/configuration#stt-options)

STT Options
-------------------------------------------------------------------------------

Copy

Ask AI

    public struct STTOptions: Sendable {
        public let language: String             // ISO 639-1 code
        public let sampleRate: Int              // Audio sample rate
        public let enableWordTimestamps: Bool   // Per-word timing
        public let enableVAD: Bool              // Voice activity detection
    }
    

| Option | Default | Description |
| --- | --- | --- |
| `language` | `"en"` | Language code (or empty for auto) |
| `sampleRate` | `16000` | Audio sample rate in Hz |
| `enableWordTimestamps` | `false` | Include word timing info |
| `enableVAD` | `true` | Filter silent segments |

[​](https://docs.runanywhere.ai/swift/configuration#tts-options)

TTS Options
-------------------------------------------------------------------------------

Copy

Ask AI

    public struct TTSOptions: Sendable {
        public let rate: Float          // Speed (0.5-2.0)
        public let pitch: Float         // Pitch (0.5-2.0)
        public let volume: Float        // Volume (0.0-1.0)
        public let language: String     // Language code
        public let sampleRate: Int      // Output sample rate
        public let audioFormat: AudioFormat
    }
    

| Option | Default | Range | Description |
| --- | --- | --- | --- |
| `rate` | `1.0` | 0.5-2.0 | Playback speed |
| `pitch` | `1.0` | 0.5-2.0 | Voice pitch |
| `volume` | `1.0` | 0.0-1.0 | Audio volume |
| `sampleRate` | `22050` | Hz  | Output sample rate |
| `audioFormat` | `.wav` | Enum | Output format |

[​](https://docs.runanywhere.ai/swift/configuration#vad-configuration)

VAD Configuration
-------------------------------------------------------------------------------------------

Copy

Ask AI

    public struct VADConfiguration: Sendable {
        public let sampleRate: Int              // Audio sample rate
        public let frameLength: Double          // Frame duration (seconds)
        public let energyThreshold: Double      // Detection sensitivity
    }
    

| Option | Default | Description |
| --- | --- | --- |
| `sampleRate` | `16000` | Audio sample rate |
| `frameLength` | `0.032` | Frame length (32ms) |
| `energyThreshold` | `0.5` | Sensitivity (0=sensitive, 1=strict) |

[​](https://docs.runanywhere.ai/swift/configuration#memory-considerations)

Memory Considerations
---------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/configuration#model-memory-usage)

Model Memory Usage

| Model Type | Typical Memory | Notes |
| --- | --- | --- |
| LLM Q4 (1B params) | 1-2 GB | Suitable for all devices |
| LLM Q4 (3B params) | 3-4 GB | iPhone Pro / iPad |
| LLM Q4 (7B params) | 6-8 GB | M1+ Macs only |
| STT (Whisper Base) | ~150 MB | Universal |
| TTS (Piper) | 50-100 MB | Universal |

### 

[​](https://docs.runanywhere.ai/swift/configuration#memory-management)

Memory Management

Copy

Ask AI

    // Unload models when not needed
    try await RunAnywhere.unloadModel()
    try await RunAnywhere.unloadSTTModel()
    try await RunAnywhere.unloadTTSVoice()
    
    // Check storage
    let storage = await RunAnywhere.getStorageInfo()
    print("Available: \(storage.availableBytes / 1_000_000_000) GB")
    
    // Clean temporary files
    try await RunAnywhere.cleanTempFiles()
    

[​](https://docs.runanywhere.ai/swift/configuration#logging-configuration)

Logging Configuration
---------------------------------------------------------------------------------------------------

Copy

Ask AI

    // Set log level
    RunAnywhere.setLogLevel(.debug)  // .debug | .info | .warning | .error | .fault
    
    // Enable local logging (Pulse integration)
    RunAnywhere.configureLocalLogging(enabled: true)
    
    // Enable verbose debug mode
    RunAnywhere.setDebugMode(true)
    
    // Flush logs
    await RunAnywhere.flushAll()
    

### 

[​](https://docs.runanywhere.ai/swift/configuration#log-levels)

Log Levels

| Level | Description |
| --- | --- |
| `.debug` | Detailed information for debugging |
| `.info` | General operational information |
| `.warning` | Potential issues that don’t prevent operation |
| `.error` | Errors that affect specific operations |
| `.fault` | Critical errors indicating serious problems |

[​](https://docs.runanywhere.ai/swift/configuration#sdk-state)

SDK State
---------------------------------------------------------------------------

Copy

Ask AI

    // Check initialization
    let isInitialized = RunAnywhere.isSDKInitialized
    let servicesReady = RunAnywhere.areServicesReady
    let isActive = RunAnywhere.isActive
    
    // Version info
    let version = RunAnywhere.version
    
    // Authentication
    let isAuthenticated = RunAnywhere.isAuthenticated
    let userId = RunAnywhere.getUserId()
    let orgId = RunAnywhere.getOrganizationId()
    
    // Device
    let deviceId = RunAnywhere.deviceId
    let isRegistered = RunAnywhere.isDeviceRegistered()
    

[​](https://docs.runanywhere.ai/swift/configuration#environment-specific-settings)

Environment-Specific Settings
-------------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    func configureForEnvironment() {
        #if DEBUG
        try? RunAnywhere.initialize(
            environment: .development
        )
        RunAnywhere.setLogLevel(.debug)
        RunAnywhere.setDebugMode(true)
        #else
        try? RunAnywhere.initialize(
            apiKey: ProcessInfo.processInfo.environment["RUNANYWHERE_API_KEY"],
            baseURL: "https://api.runanywhere.ai",
            environment: .production
        )
        RunAnywhere.setLogLevel(.warning)
        #endif
    }
    

[​](https://docs.runanywhere.ai/swift/configuration#reset-testing)

Reset (Testing)
-------------------------------------------------------------------------------------

Copy

Ask AI

    // Reset SDK state for testing
    RunAnywhere.reset()
    

Only use `reset()` in development/testing. It clears all SDK state including authentication.

[Error Handling\
--------------\
\
Handle SDK errors →](https://docs.runanywhere.ai/swift/error-handling)
[Best Practices\
--------------\
\
Performance tips →](https://docs.runanywhere.ai/swift/best-practices)

[Voice Agent](https://docs.runanywhere.ai/swift/voice-agent)
[Error Handling](https://docs.runanywhere.ai/swift/error-handling)

⌘I

---

# chat() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/llm/chat#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

chat()

On this page

*   [Basic Usage](https://docs.runanywhere.ai/swift/llm/chat#basic-usage)
    
*   [Method Signature](https://docs.runanywhere.ai/swift/llm/chat#method-signature)
    
*   [Parameters](https://docs.runanywhere.ai/swift/llm/chat#parameters)
    
*   [Returns](https://docs.runanywhere.ai/swift/llm/chat#returns)
    
*   [Throws](https://docs.runanywhere.ai/swift/llm/chat#throws)
    
*   [Examples](https://docs.runanywhere.ai/swift/llm/chat#examples)
    
*   [Simple Q&A](https://docs.runanywhere.ai/swift/llm/chat#simple-q%26a)
    
*   [Conversation Flow](https://docs.runanywhere.ai/swift/llm/chat#conversation-flow)
    
*   [Error Handling](https://docs.runanywhere.ai/swift/llm/chat#error-handling)
    
*   [When to Use chat() vs generate()](https://docs.runanywhere.ai/swift/llm/chat#when-to-use-chat-vs-generate)
    
*   [SwiftUI Example](https://docs.runanywhere.ai/swift/llm/chat#swiftui-example)
    

The `chat()` method provides the simplest way to generate text responses. It returns just the text string, making it perfect for quick interactions.

[​](https://docs.runanywhere.ai/swift/llm/chat#basic-usage)

Basic Usage
--------------------------------------------------------------------------

Copy

Ask AI

    let response = try await RunAnywhere.chat("What is the capital of France?")
    print(response)  // "The capital of France is Paris."
    

[​](https://docs.runanywhere.ai/swift/llm/chat#method-signature)

Method Signature
------------------------------------------------------------------------------------

Copy

Ask AI

    public static func chat(_ prompt: String) async throws -> String
    

### 

[​](https://docs.runanywhere.ai/swift/llm/chat#parameters)

Parameters

| Parameter | Type | Description |
| --- | --- | --- |
| `prompt` | `String` | The text prompt to respond to |

### 

[​](https://docs.runanywhere.ai/swift/llm/chat#returns)

Returns

A `String` containing the generated response text.

### 

[​](https://docs.runanywhere.ai/swift/llm/chat#throws)

Throws

*   `SDKError` if the SDK is not initialized
*   `SDKError` if no LLM model is loaded
*   `SDKError` if generation fails

[​](https://docs.runanywhere.ai/swift/llm/chat#examples)

Examples
--------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/llm/chat#simple-q&a)

Simple Q&A

Copy

Ask AI

    // Simple question
    let answer = try await RunAnywhere.chat("What's 2 + 2?")
    print(answer)  // "4" or "The answer is 4."
    
    // More complex question
    let explanation = try await RunAnywhere.chat(
        "Explain the difference between let and var in Swift"
    )
    print(explanation)
    

### 

[​](https://docs.runanywhere.ai/swift/llm/chat#conversation-flow)

Conversation Flow

Copy

Ask AI

    // While chat() doesn't maintain conversation history,
    // you can build context manually
    var context = ""
    
    func ask(_ question: String) async throws -> String {
        let fullPrompt = context.isEmpty
            ? question
            : "\(context)\n\nUser: \(question)"
    
        let response = try await RunAnywhere.chat(fullPrompt)
    
        // Append to context for follow-up
        context += "\nUser: \(question)\nAssistant: \(response)"
    
        return response
    }
    
    let response1 = try await ask("My name is Alice")
    let response2 = try await ask("What's my name?")  // Will remember "Alice"
    

### 

[​](https://docs.runanywhere.ai/swift/llm/chat#error-handling)

Error Handling

Copy

Ask AI

    do {
        let response = try await RunAnywhere.chat("Hello!")
        print(response)
    } catch let error as SDKError {
        switch error.code {
        case .notInitialized:
            print("SDK not initialized. Call RunAnywhere.initialize() first.")
        case .modelNotFound:
            print("No model loaded. Call RunAnywhere.loadModel() first.")
        case .generationFailed:
            print("Generation failed: \(error.message)")
        default:
            print("Error: \(error.localizedDescription)")
        }
    }
    

[​](https://docs.runanywhere.ai/swift/llm/chat#when-to-use-chat-vs-generate)

When to Use chat() vs generate()
----------------------------------------------------------------------------------------------------------------

| Use Case | Method |
| --- | --- |
| Quick responses | `chat()` |
| Need performance metrics | `generate()` |
| Custom generation options | `generate()` |
| Real-time token display | `generateStream()` |
| Simple prototyping | `chat()` |
| Production with analytics | `generate()` |

[​](https://docs.runanywhere.ai/swift/llm/chat#swiftui-example)

SwiftUI Example
----------------------------------------------------------------------------------

Copy

Ask AI

    struct ChatView: View {
        @State private var input = ""
        @State private var messages: [(String, Bool)] = []  // (text, isUser)
        @State private var isLoading = false
    
        var body: some View {
            VStack {
                // Messages
                ScrollView {
                    ForEach(Array(messages.enumerated()), id: \.offset) { _, message in
                        HStack {
                            if message.1 { Spacer() }
                            Text(message.0)
                                .padding()
                                .background(message.1 ? Color.blue : Color.gray.opacity(0.2))
                                .foregroundColor(message.1 ? .white : .primary)
                                .cornerRadius(12)
                            if !message.1 { Spacer() }
                        }
                    }
                }
    
                // Input
                HStack {
                    TextField("Message...", text: $input)
                        .textFieldStyle(.roundedBorder)
    
                    Button(action: sendMessage) {
                        Image(systemName: "arrow.up.circle.fill")
                            .font(.title)
                    }
                    .disabled(input.isEmpty || isLoading)
                }
                .padding()
            }
        }
    
        func sendMessage() {
            let userMessage = input
            messages.append((userMessage, true))
            input = ""
            isLoading = true
    
            Task {
                do {
                    let response = try await RunAnywhere.chat(userMessage)
                    await MainActor.run {
                        messages.append((response, false))
                        isLoading = false
                    }
                } catch {
                    await MainActor.run {
                        messages.append(("Error: \(error.localizedDescription)", false))
                        isLoading = false
                    }
                }
            }
        }
    }
    

[generate()\
----------\
\
For detailed metrics and custom options, use generate() →](https://docs.runanywhere.ai/swift/llm/generate)

[Quick Start](https://docs.runanywhere.ai/swift/quick-start)
[generate()](https://docs.runanywhere.ai/swift/llm/generate)

⌘I

---

# System Prompts - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/llm/system-prompts#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

System Prompts

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/llm/system-prompts#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/react-native/llm/system-prompts#basic-usage)
    
*   [Best Practices](https://docs.runanywhere.ai/react-native/llm/system-prompts#best-practices)
    
*   [Be Specific](https://docs.runanywhere.ai/react-native/llm/system-prompts#be-specific)
    
*   [Define Output Format](https://docs.runanywhere.ai/react-native/llm/system-prompts#define-output-format)
    
*   [Set Constraints](https://docs.runanywhere.ai/react-native/llm/system-prompts#set-constraints)
    
*   [Example Prompts](https://docs.runanywhere.ai/react-native/llm/system-prompts#example-prompts)
    
*   [Coding Assistant](https://docs.runanywhere.ai/react-native/llm/system-prompts#coding-assistant)
    
*   [Creative Writer](https://docs.runanywhere.ai/react-native/llm/system-prompts#creative-writer)
    
*   [Data Analyst](https://docs.runanywhere.ai/react-native/llm/system-prompts#data-analyst)
    
*   [Language Tutor](https://docs.runanywhere.ai/react-native/llm/system-prompts#language-tutor)
    
*   [Technical Support](https://docs.runanywhere.ai/react-native/llm/system-prompts#technical-support)
    
*   [Multi-Turn Conversations](https://docs.runanywhere.ai/react-native/llm/system-prompts#multi-turn-conversations)
    
*   [Prompt Templates](https://docs.runanywhere.ai/react-native/llm/system-prompts#prompt-templates)
    
*   [Tips](https://docs.runanywhere.ai/react-native/llm/system-prompts#tips)
    
*   [Related](https://docs.runanywhere.ai/react-native/llm/system-prompts#related)
    

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#overview)

Overview
-------------------------------------------------------------------------------------

System prompts define how the AI behaves, its personality, constraints, and the format of responses. They’re set once and apply to all subsequent generations until changed.

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#basic-usage)

Basic Usage
-------------------------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    const result = await RunAnywhere.generate('What is the best programming language?', {
      maxTokens: 200,
      systemPrompt:
        'You are a helpful coding assistant. Be concise and practical. Answer in 2-3 sentences.',
    })
    

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#best-practices)

Best Practices
-------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#be-specific)

Be Specific

Copy

Ask AI

    // ❌ Too vague
    const vague = {
      systemPrompt: 'Be helpful.',
    }
    
    // ✅ Specific and clear
    const specific = {
      systemPrompt: `You are a senior software engineer helping junior developers.
    - Give practical, production-ready code examples
    - Explain your reasoning briefly
    - Mention potential pitfalls when relevant`,
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#define-output-format)

Define Output Format

Copy

Ask AI

    // JSON output
    const jsonPrompt = {
      systemPrompt: `You are a data extraction assistant.
    Always respond with valid JSON in this format:
    {
      "entities": ["entity1", "entity2"],
      "sentiment": "positive" | "negative" | "neutral",
      "confidence": 0.0-1.0
    }
    Do not include any text outside the JSON.`,
    }
    
    // List format
    const listPrompt = {
      systemPrompt: `You are a task breakdown assistant.
    Always respond with a numbered list of actionable steps.
    Keep each step to one sentence.
    Include 3-7 steps per response.`,
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#set-constraints)

Set Constraints

Copy

Ask AI

    const constrained = {
      systemPrompt: `You are a customer support assistant for TechCorp.
    Rules:
    - Only answer questions about our products
    - Never discuss competitors
    - If you don't know something, say "I'll connect you with a human agent"
    - Keep responses under 100 words
    - Be friendly but professional`,
    }
    

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#example-prompts)

Example Prompts
---------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#coding-assistant)

Coding Assistant

Copy

Ask AI

    const codingAssistant = await RunAnywhere.generate('How do I read a file in Python?', {
      maxTokens: 300,
      systemPrompt: `You are an expert programmer helping developers.
    - Provide working code examples
    - Use modern best practices
    - Include error handling
    - Add brief comments explaining key parts
    - Mention the Python version if relevant`,
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#creative-writer)

Creative Writer

Copy

Ask AI

    const creativeWriter = await RunAnywhere.generate('Write a story opening', {
      maxTokens: 200,
      temperature: 1.0,
      systemPrompt: `You are a creative fiction writer.
    - Use vivid, sensory descriptions
    - Create engaging hooks
    - Vary sentence length and structure
    - Show, don't tell`,
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#data-analyst)

Data Analyst

Copy

Ask AI

    const dataAnalyst = await RunAnywhere.generate(
      'Analyze these sales numbers: Q1: $10M, Q2: $12M, Q3: $9M, Q4: $15M',
      {
        maxTokens: 250,
        temperature: 0.3,
        systemPrompt: `You are a business analyst.
    - Focus on trends and patterns
    - Calculate relevant metrics (growth rate, average, etc.)
    - Provide actionable insights
    - Use bullet points for clarity`,
      }
    )
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#language-tutor)

Language Tutor

Copy

Ask AI

    const languageTutor = await RunAnywhere.generate(
      'How do I say "Where is the train station?" in Japanese?',
      {
        maxTokens: 150,
        systemPrompt: `You are a language learning assistant.
    - Provide the translation with pronunciation guide
    - Break down the grammar
    - Give a literal translation
    - Include one related useful phrase`,
      }
    )
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#technical-support)

Technical Support

Copy

Ask AI

    const techSupport = await RunAnywhere.generate(
      'My app keeps crashing when I click the submit button',
      {
        maxTokens: 300,
        systemPrompt: `You are a technical support specialist.
    - Ask clarifying questions if needed
    - Provide step-by-step troubleshooting
    - Start with the most common solutions
    - Explain why each step might help`,
      }
    )
    

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#multi-turn-conversations)

Multi-Turn Conversations
---------------------------------------------------------------------------------------------------------------------

For multi-turn conversations, you can include conversation history in the prompt:

Copy

Ask AI

    interface Message {
      role: 'user' | 'assistant'
      content: string
    }
    
    async function chat(messages: Message[], newMessage: string): Promise<string> {
      // Build conversation history
      const history = messages
        .map((m) => `${m.role === 'user' ? 'User' : 'Assistant'}: ${m.content}`)
        .join('\n')
    
      const prompt = history ? `${history}\nUser: ${newMessage}\nAssistant:` : newMessage
    
      const result = await RunAnywhere.generate(prompt, {
        maxTokens: 200,
        systemPrompt: `You are a helpful assistant engaged in a conversation.
    Respond naturally and maintain context from previous messages.
    Be concise but complete.`,
      })
    
      return result.text
    }
    
    // Usage
    const messages: Message[] = []
    
    const response1 = await chat(messages, "What's TypeScript?")
    messages.push({ role: 'user', content: "What's TypeScript?" })
    messages.push({ role: 'assistant', content: response1 })
    
    const response2 = await chat(messages, 'How does it compare to JavaScript?')
    // Response will have context from the previous exchange
    

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#prompt-templates)

Prompt Templates
-----------------------------------------------------------------------------------------------------

Create reusable prompt templates:

Copy

Ask AI

    const promptTemplates = {
      codeReview: (language: string) => `You are a senior ${language} developer.
    Review code for:
    - Bugs and potential issues
    - Performance problems
    - Best practice violations
    - Security concerns
    Provide specific suggestions with examples.`,
    
      summarize: (style: 'brief' | 'detailed') => `You are a summarization assistant.
    Create a ${style === 'brief' ? 'one-paragraph' : 'comprehensive'} summary.
    ${style === 'detailed' ? 'Include key points as bullet points.' : 'Focus on the main takeaway.'}`,
    
      translate: (targetLanguage: string) => `You are a professional translator.
    Translate text to ${targetLanguage}.
    - Preserve the original meaning and tone
    - Use natural expressions in the target language
    - Keep formatting (lists, paragraphs) intact`,
    }
    
    // Usage
    const review = await RunAnywhere.generate(codeSnippet, {
      maxTokens: 400,
      systemPrompt: promptTemplates.codeReview('TypeScript'),
    })
    

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#tips)

Tips
-----------------------------------------------------------------------------

**Keep system prompts concise.** Long system prompts consume tokens and slow down generation. Focus on the most important instructions.

**Test with different temperatures.** Lower temperatures (0.1-0.3) work better for factual/structured outputs. Higher temperatures (0.7-1.2) work better for creative tasks.

System prompts are suggestions, not guarantees. Models may not always follow instructions perfectly, especially with complex or conflicting rules. Test thoroughly.

[​](https://docs.runanywhere.ai/react-native/llm/system-prompts#related)

Related
-----------------------------------------------------------------------------------

[Generate\
--------\
\
Full generation API reference](https://docs.runanywhere.ai/react-native/llm/generate)
[Streaming\
---------\
\
Real-time token streaming](https://docs.runanywhere.ai/react-native/llm/stream)

[generateStream()](https://docs.runanywhere.ai/react-native/llm/stream)
[transcribe()](https://docs.runanywhere.ai/react-native/stt/transcribe)

⌘I

---

# Quick Start - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/quick-start#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Swift SDK

Quick Start

On this page

*   [Complete Example](https://docs.runanywhere.ai/swift/quick-start#complete-example)
    
*   [Step 1: Initialize the SDK](https://docs.runanywhere.ai/swift/quick-start#step-1-initialize-the-sdk)
    
*   [Environment Options](https://docs.runanywhere.ai/swift/quick-start#environment-options)
    
*   [Step 2: Register Modules](https://docs.runanywhere.ai/swift/quick-start#step-2-register-modules)
    
*   [Step 3: Generate Text](https://docs.runanywhere.ai/swift/quick-start#step-3-generate-text)
    
*   [Simple Chat](https://docs.runanywhere.ai/swift/quick-start#simple-chat)
    
*   [Full Generation with Metrics](https://docs.runanywhere.ai/swift/quick-start#full-generation-with-metrics)
    
*   [Step 4: Load Models](https://docs.runanywhere.ai/swift/quick-start#step-4-load-models)
    
*   [Complete SwiftUI Example](https://docs.runanywhere.ai/swift/quick-start#complete-swiftui-example)
    
*   [Streaming Example](https://docs.runanywhere.ai/swift/quick-start#streaming-example)
    
*   [What’s Next?](https://docs.runanywhere.ai/swift/quick-start#what%E2%80%99s-next)
    

[​](https://docs.runanywhere.ai/swift/quick-start#complete-example)

Complete Example
---------------------------------------------------------------------------------------

Here’s a complete example to get you started with on-device text generation:

Copy

Ask AI

    import RunAnywhere
    import LlamaCPPRuntime
    
    @main
    struct MyApp: App {
        init() {
            Task { @MainActor in
                // Register the LlamaCPP module for LLM support
                LlamaCPP.register()
    
                // Initialize the SDK
                do {
                    try RunAnywhere.initialize(
                        apiKey: "<YOUR_API_KEY>",
                        baseURL: "https://api.runanywhere.ai",
                        environment: .production
                    )
                } catch {
                    print("SDK initialization failed: \(error)")
                }
            }
        }
    
        var body: some Scene {
            WindowGroup {
                ContentView()
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/quick-start#step-1-initialize-the-sdk)

Step 1: Initialize the SDK
----------------------------------------------------------------------------------------------------------

Initialize the SDK once at app launch:

Copy

Ask AI

    try RunAnywhere.initialize(
        apiKey: "<YOUR_API_KEY>",       // Required for production
        baseURL: "https://api.runanywhere.ai",
        environment: .production        // .development | .staging | .production
    )
    

### 

[​](https://docs.runanywhere.ai/swift/quick-start#environment-options)

Environment Options

| Environment | Log Level | Description |
| --- | --- | --- |
| `.development` | Debug | Verbose logging, mock services, local analytics |
| `.staging` | Info | Testing with real services |
| `.production` | Warning | Minimal logging, full authentication, telemetry |

[​](https://docs.runanywhere.ai/swift/quick-start#step-2-register-modules)

Step 2: Register Modules
------------------------------------------------------------------------------------------------------

Register modules at app startup before using their capabilities:

Copy

Ask AI

    import RunAnywhere
    import LlamaCPPRuntime
    import ONNXRuntime
    
    @MainActor
    func setupSDK() {
        LlamaCPP.register()   // LLM (priority: 100)
        ONNX.register()       // STT + TTS + VAD (priority: 100)
    }
    

[​](https://docs.runanywhere.ai/swift/quick-start#step-3-generate-text)

Step 3: Generate Text
------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/quick-start#simple-chat)

Simple Chat

Copy

Ask AI

    let response = try await RunAnywhere.chat("What is the capital of France?")
    print(response)  // "The capital of France is Paris."
    

### 

[​](https://docs.runanywhere.ai/swift/quick-start#full-generation-with-metrics)

Full Generation with Metrics

Copy

Ask AI

    let result = try await RunAnywhere.generate(
        "Explain quantum computing in simple terms",
        options: LLMGenerationOptions(
            maxTokens: 200,
            temperature: 0.7
        )
    )
    
    print("Response: \(result.text)")
    print("Tokens used: \(result.tokensUsed)")
    print("Speed: \(result.tokensPerSecond) tok/s")
    print("Latency: \(result.latencyMs)ms")
    

[​](https://docs.runanywhere.ai/swift/quick-start#step-4-load-models)

Step 4: Load Models
--------------------------------------------------------------------------------------------

Copy

Ask AI

    // Load an LLM model by ID
    try await RunAnywhere.loadModel("llama-3.2-1b-instruct-q4")
    
    // Check if model is loaded
    let isLoaded = await RunAnywhere.isModelLoaded
    print("Model loaded: \(isLoaded)")
    
    // Get current model
    let modelId = await RunAnywhere.getCurrentModelId()
    

[​](https://docs.runanywhere.ai/swift/quick-start#complete-swiftui-example)

Complete SwiftUI Example
-------------------------------------------------------------------------------------------------------

Here’s a full SwiftUI view demonstrating text generation:

Copy

Ask AI

    import SwiftUI
    import RunAnywhere
    
    struct ContentView: View {
        @State private var prompt = ""
        @State private var response = ""
        @State private var isLoading = false
        @State private var isModelLoaded = false
    
        var body: some View {
            VStack(spacing: 20) {
                // Model status
                HStack {
                    Circle()
                        .fill(isModelLoaded ? .green : .gray)
                        .frame(width: 10, height: 10)
                    Text(isModelLoaded ? "Model Ready" : "Model Not Loaded")
                        .font(.caption)
                }
    
                // Input
                TextField("Enter your prompt...", text: $prompt)
                    .textFieldStyle(.roundedBorder)
                    .padding(.horizontal)
    
                // Generate button
                Button(action: { Task { await generate() } }) {
                    HStack {
                        if isLoading {
                            ProgressView()
                                .scaleEffect(0.8)
                        }
                        Text(isLoading ? "Generating..." : "Generate")
                    }
                    .frame(maxWidth: .infinity)
                    .padding()
                    .background(Color.blue)
                    .foregroundColor(.white)
                    .cornerRadius(10)
                }
                .disabled(isLoading || prompt.isEmpty || !isModelLoaded)
                .padding(.horizontal)
    
                // Response
                ScrollView {
                    Text(response)
                        .padding()
                        .frame(maxWidth: .infinity, alignment: .leading)
                }
                .background(Color.gray.opacity(0.1))
                .cornerRadius(10)
                .padding(.horizontal)
            }
            .padding()
            .task {
                await loadModel()
            }
        }
    
        func loadModel() async {
            do {
                try await RunAnywhere.loadModel("llama-3.2-1b-instruct-q4")
                isModelLoaded = true
            } catch {
                response = "Failed to load model: \(error.localizedDescription)"
            }
        }
    
        func generate() async {
            isLoading = true
            defer { isLoading = false }
    
            do {
                let result = try await RunAnywhere.generate(
                    prompt,
                    options: LLMGenerationOptions(
                        maxTokens: 200,
                        temperature: 0.7
                    )
                )
                response = result.text
            } catch {
                response = "Error: \(error.localizedDescription)"
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/quick-start#streaming-example)

Streaming Example
-----------------------------------------------------------------------------------------

For a more responsive UI, use streaming generation:

Copy

Ask AI

    func generateStreaming() async {
        isLoading = true
        response = ""
    
        do {
            let result = try await RunAnywhere.generateStream(
                prompt,
                options: LLMGenerationOptions(maxTokens: 500)
            )
    
            // Display tokens as they arrive
            for try await token in result.stream {
                await MainActor.run {
                    response += token
                }
            }
    
            // Get final metrics
            let metrics = try await result.result.value
            print("Generated \(metrics.tokensUsed) tokens at \(metrics.tokensPerSecond) tok/s")
    
        } catch {
            response = "Error: \(error.localizedDescription)"
        }
    
        isLoading = false
    }
    

[​](https://docs.runanywhere.ai/swift/quick-start#what%E2%80%99s-next)

What’s Next?
--------------------------------------------------------------------------------------

[LLM Guide\
---------\
\
Learn about text generation options](https://docs.runanywhere.ai/swift/llm/chat)
[Speech-to-Text\
--------------\
\
Add voice transcription](https://docs.runanywhere.ai/swift/stt/transcribe)
[Text-to-Speech\
--------------\
\
Generate speech from text](https://docs.runanywhere.ai/swift/tts/synthesize)
[Voice Agent\
-----------\
\
Build complete voice experiences](https://docs.runanywhere.ai/swift/voice-agent)

[Installation](https://docs.runanywhere.ai/swift/installation)
[chat()](https://docs.runanywhere.ai/swift/llm/chat)

⌘I

---

# Best Practices - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/best-practices#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

React Native SDK

Best Practices

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/best-practices#overview)
    
*   [Model Selection](https://docs.runanywhere.ai/react-native/best-practices#model-selection)
    
*   [Choose the Right Model Size](https://docs.runanywhere.ai/react-native/best-practices#choose-the-right-model-size)
    
*   [Quantization Trade-offs](https://docs.runanywhere.ai/react-native/best-practices#quantization-trade-offs)
    
*   [Memory Management](https://docs.runanywhere.ai/react-native/best-practices#memory-management)
    
*   [Unload Unused Models](https://docs.runanywhere.ai/react-native/best-practices#unload-unused-models)
    
*   [Handle App Lifecycle](https://docs.runanywhere.ai/react-native/best-practices#handle-app-lifecycle)
    
*   [Check Memory Before Loading](https://docs.runanywhere.ai/react-native/best-practices#check-memory-before-loading)
    
*   [Performance Optimization](https://docs.runanywhere.ai/react-native/best-practices#performance-optimization)
    
*   [Use Streaming for Better UX](https://docs.runanywhere.ai/react-native/best-practices#use-streaming-for-better-ux)
    
*   [Limit Token Generation](https://docs.runanywhere.ai/react-native/best-practices#limit-token-generation)
    
*   [Pre-Download Models](https://docs.runanywhere.ai/react-native/best-practices#pre-download-models)
    
*   [Error Handling](https://docs.runanywhere.ai/react-native/best-practices#error-handling)
    
*   [Always Handle Errors Gracefully](https://docs.runanywhere.ai/react-native/best-practices#always-handle-errors-gracefully)
    
*   [Provide User Feedback](https://docs.runanywhere.ai/react-native/best-practices#provide-user-feedback)
    
*   [User Experience](https://docs.runanywhere.ai/react-native/best-practices#user-experience)
    
*   [Show Progress During Downloads](https://docs.runanywhere.ai/react-native/best-practices#show-progress-during-downloads)
    
*   [Add Typing Indicators](https://docs.runanywhere.ai/react-native/best-practices#add-typing-indicators)
    
*   [Graceful Degradation](https://docs.runanywhere.ai/react-native/best-practices#graceful-degradation)
    
*   [Security & Privacy](https://docs.runanywhere.ai/react-native/best-practices#security-%26-privacy)
    
*   [Never Log Sensitive Data](https://docs.runanywhere.ai/react-native/best-practices#never-log-sensitive-data)
    
*   [Use Development Mode Appropriately](https://docs.runanywhere.ai/react-native/best-practices#use-development-mode-appropriately)
    
*   [Disable Telemetry If Needed](https://docs.runanywhere.ai/react-native/best-practices#disable-telemetry-if-needed)
    
*   [Testing](https://docs.runanywhere.ai/react-native/best-practices#testing)
    
*   [Use Smaller Models for Testing](https://docs.runanywhere.ai/react-native/best-practices#use-smaller-models-for-testing)
    
*   [Mock for Unit Tests](https://docs.runanywhere.ai/react-native/best-practices#mock-for-unit-tests)
    
*   [Summary Checklist](https://docs.runanywhere.ai/react-native/best-practices#summary-checklist)
    
*   [Related](https://docs.runanywhere.ai/react-native/best-practices#related)
    

[​](https://docs.runanywhere.ai/react-native/best-practices#overview)

Overview
---------------------------------------------------------------------------------

This guide covers best practices for building performant, reliable, and user-friendly AI applications with the RunAnywhere SDK.

[​](https://docs.runanywhere.ai/react-native/best-practices#model-selection)

Model Selection
-----------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#choose-the-right-model-size)

Choose the Right Model Size

| Model Size | RAM Required | Use Case | Speed |
| --- | --- | --- | --- |
| 360M–500M (Q8) | ~500MB | Quick responses, chat | Very Fast |
| 1B–3B (Q4/Q6) | 1–2GB | Balanced quality/speed | Fast |
| 7B (Q4) | 4–5GB | High quality | Slower |

Copy

Ask AI

    // For chat applications - use smaller, faster models
    await LlamaCPP.addModel({
      id: 'smollm2-360m',
      name: 'SmolLM2 360M',
      url: 'https://huggingface.co/.../SmolLM2-360M.Q8_0.gguf',
      memoryRequirement: 500_000_000,
    })
    
    // For quality-critical tasks - use larger models
    await LlamaCPP.addModel({
      id: 'qwen-1.5b',
      name: 'Qwen 1.5B',
      url: 'https://huggingface.co/.../qwen-1.5b-q4_k_m.gguf',
      memoryRequirement: 1_500_000_000,
    })
    

Start with smaller models during development for faster iteration. Switch to larger models when quality becomes critical.

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#quantization-trade-offs)

Quantization Trade-offs

| Quantization | Quality | Size | Speed |
| --- | --- | --- | --- |
| Q8\_0 | Best | Largest | Slower |
| Q6\_K | Great | Large | Fast |
| Q4\_K\_M | Good | Medium | Faster |
| Q4\_0 | Acceptable | Small | Fastest |

[​](https://docs.runanywhere.ai/react-native/best-practices#memory-management)

Memory Management
---------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#unload-unused-models)

Unload Unused Models

Copy

Ask AI

    // Unload LLM when not in use
    await RunAnywhere.unloadModel()
    
    // Unload STT when not needed
    await RunAnywhere.unloadSTTModel()
    
    // Unload TTS when done
    await RunAnywhere.unloadTTSModel()
    

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#handle-app-lifecycle)

Handle App Lifecycle

Copy

Ask AI

    import { useEffect } from 'react'
    import { AppState, AppStateStatus } from 'react-native'
    import { RunAnywhere } from '@runanywhere/core'
    
    function useModelLifecycle(modelId: string) {
      useEffect(() => {
        const subscription = AppState.addEventListener('change', (state: AppStateStatus) => {
          if (state === 'background') {
            // Free memory when app backgrounds
            RunAnywhere.unloadModel()
          } else if (state === 'active') {
            // Optionally reload when app returns
            // This depends on your UX requirements
          }
        })
    
        return () => subscription.remove()
      }, [modelId])
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#check-memory-before-loading)

Check Memory Before Loading

Copy

Ask AI

    async function safeLoadModel(modelId: string): Promise<boolean> {
      const modelInfo = await RunAnywhere.getModelInfo(modelId)
      const storage = await RunAnywhere.getStorageInfo()
    
      if (!modelInfo?.memoryRequired) {
        return false
      }
    
      // Check if we have enough free memory (with 20% buffer)
      const requiredWithBuffer = modelInfo.memoryRequired * 1.2
    
      if (storage.freeSpace < requiredWithBuffer) {
        console.warn('Insufficient memory for model')
        return false
      }
    
      await RunAnywhere.loadModel(modelInfo.localPath!)
      return true
    }
    

[​](https://docs.runanywhere.ai/react-native/best-practices#performance-optimization)

Performance Optimization
-----------------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#use-streaming-for-better-ux)

Use Streaming for Better UX

Copy

Ask AI

    // ❌ Bad: User waits for entire response
    const result = await RunAnywhere.generate(prompt, { maxTokens: 500 })
    setResponse(result.text)
    
    // ✅ Good: User sees response as it's generated
    const stream = await RunAnywhere.generateStream(prompt, { maxTokens: 500 })
    for await (const token of stream.stream) {
      setResponse((prev) => prev + token)
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#limit-token-generation)

Limit Token Generation

Copy

Ask AI

    // For quick responses
    const quick = await RunAnywhere.generate(prompt, {
      maxTokens: 100, // Short responses
      temperature: 0.5, // Faster sampling
    })
    
    // For detailed responses
    const detailed = await RunAnywhere.generate(prompt, {
      maxTokens: 500,
      temperature: 0.7,
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#pre-download-models)

Pre-Download Models

Download models during onboarding for a better user experience:

OnboardingScreen.tsx

Copy

Ask AI

    async function downloadModels(onProgress: (percent: number) => void) {
      const models = ['smollm2-360m', 'whisper-tiny-en', 'piper-en-lessac']
    
      for (let i = 0; i < models.length; i++) {
        await RunAnywhere.downloadModel(models[i], (progress) => {
          const overallProgress = (i + progress.progress) / models.length
          onProgress(overallProgress * 100)
        })
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/best-practices#error-handling)

Error Handling
---------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#always-handle-errors-gracefully)

Always Handle Errors Gracefully

Copy

Ask AI

    async function generateSafely(prompt: string): Promise<string> {
      try {
        const result = await RunAnywhere.generate(prompt)
        return result.text
      } catch (error) {
        if (isSDKError(error)) {
          switch (error.code) {
            case SDKErrorCode.modelNotLoaded:
              // Try to load model and retry
              await loadDefaultModel()
              return generateSafely(prompt)
            case SDKErrorCode.insufficientMemory:
              // Use a smaller model
              return 'I need to use a smaller model. Please try again.'
            case SDKErrorCode.generationCancelled:
              return '' // Expected, don't show error
            default:
              return 'Sorry, I encountered an error. Please try again.'
          }
        }
        return 'An unexpected error occurred.'
      }
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#provide-user-feedback)

Provide User Feedback

Copy

Ask AI

    // Show loading states
    const [status, setStatus] = useState<'idle' | 'loading' | 'generating' | 'error'>('idle')
    
    async function generate() {
      setStatus('loading')
      try {
        const result = await RunAnywhere.generate(prompt)
        setStatus('idle')
        return result
      } catch (error) {
        setStatus('error')
        throw error
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/best-practices#user-experience)

User Experience
-----------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#show-progress-during-downloads)

Show Progress During Downloads

DownloadProgress.tsx

Copy

Ask AI

    import React, { useState } from 'react'
    import { View, Text, ActivityIndicator, StyleSheet } from 'react-native'
    import { RunAnywhere, DownloadState } from '@runanywhere/core'
    
    export function ModelDownloader({ modelId }: { modelId: string }) {
      const [progress, setProgress] = useState(0)
      const [status, setStatus] = useState<DownloadState>('queued')
    
      const download = async () => {
        await RunAnywhere.downloadModel(modelId, (p) => {
          setProgress(p.progress * 100)
          setStatus(p.state)
        })
      }
    
      return (
        <View style={styles.container}>
          {status === 'downloading' && (
            <>
              <ActivityIndicator />
              <Text>Downloading: {progress.toFixed(0)}%</Text>
            </>
          )}
          {status === 'extracting' && (
            <>
              <ActivityIndicator />
              <Text>Setting up model...</Text>
            </>
          )}
          {status === 'completed' && <Text>✅ Ready to use!</Text>}
        </View>
      )
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#add-typing-indicators)

Add Typing Indicators

TypingIndicator.tsx

Copy

Ask AI

    export function TypingIndicator() {
      return (
        <View style={styles.container}>
          <Text style={styles.dot}>•</Text>
          <Text style={[styles.dot, styles.dot2]}>•</Text>
          <Text style={[styles.dot, styles.dot3]}>•</Text>
        </View>
      )
    }
    
    const styles = StyleSheet.create({
      container: { flexDirection: 'row', padding: 8 },
      dot: {
        fontSize: 24,
        marginHorizontal: 2,
        opacity: 0.3,
      },
      dot2: { animationDelay: '0.2s' },
      dot3: { animationDelay: '0.4s' },
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#graceful-degradation)

Graceful Degradation

Copy

Ask AI

    async function getAIResponse(prompt: string): Promise<string> {
      // Try on-device first
      if (await RunAnywhere.isModelLoaded()) {
        try {
          return (await RunAnywhere.generate(prompt)).text
        } catch {
          // Fall through to fallback
        }
      }
    
      // Fallback to simpler response
      return "I'm still setting up. Please try again in a moment."
    }
    

[​](https://docs.runanywhere.ai/react-native/best-practices#security-&-privacy)

Security & Privacy
-----------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#never-log-sensitive-data)

Never Log Sensitive Data

Copy

Ask AI

    // ❌ Bad: Logging user prompts
    console.log('User prompt:', prompt)
    logger.debug('Generating response for:', prompt)
    
    // ✅ Good: Log only metadata
    logger.debug('Generating response', { promptLength: prompt.length })
    

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#use-development-mode-appropriately)

Use Development Mode Appropriately

Copy

Ask AI

    // Development: Full logging, no auth
    await RunAnywhere.initialize({
      environment: SDKEnvironment.Development,
    })
    
    // Production: Minimal logging, with auth
    await RunAnywhere.initialize({
      environment: SDKEnvironment.Production,
      apiKey: process.env.RUNANYWHERE_API_KEY,
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#disable-telemetry-if-needed)

Disable Telemetry If Needed

Copy

Ask AI

    // Telemetry is only in Production mode
    // Use Development mode to disable all telemetry
    await RunAnywhere.initialize({
      environment: SDKEnvironment.Development, // No telemetry
    })
    

[​](https://docs.runanywhere.ai/react-native/best-practices#testing)

Testing
-------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#use-smaller-models-for-testing)

Use Smaller Models for Testing

Copy

Ask AI

    // In tests, use the smallest available model
    const testModel = 'smollm2-360m' // Fast for CI/CD
    
    beforeAll(async () => {
      await RunAnywhere.initialize({ environment: SDKEnvironment.Development })
      LlamaCPP.register()
      // Download and load small model for tests
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/best-practices#mock-for-unit-tests)

Mock for Unit Tests

Copy

Ask AI

    // Create a mock for unit tests
    jest.mock('@runanywhere/core', () => ({
      RunAnywhere: {
        generate: jest.fn().mockResolvedValue({
          text: 'Mock response',
          tokensUsed: 10,
          latencyMs: 100,
        }),
        chat: jest.fn().mockResolvedValue('Mock response'),
      },
    }))
    

[​](https://docs.runanywhere.ai/react-native/best-practices#summary-checklist)

Summary Checklist
---------------------------------------------------------------------------------------------------

Choose appropriate model size for your use case

Use streaming for better perceived performance

Unload models when not in use

Handle all error cases gracefully

Show progress during downloads and loading

Pre-download models during onboarding

Test on actual target devices

Use Development mode to iterate quickly

Never log sensitive user data

Provide clear feedback for all operations

[​](https://docs.runanywhere.ai/react-native/best-practices#related)

Related
-------------------------------------------------------------------------------

[Configuration\
-------------\
\
SDK configuration](https://docs.runanywhere.ai/react-native/configuration)
[Error Handling\
--------------\
\
Handle errors gracefully](https://docs.runanywhere.ai/react-native/error-handling)
[Quick Start\
-----------\
\
Getting started guide](https://docs.runanywhere.ai/react-native/quick-start)

[Error Handling](https://docs.runanywhere.ai/react-native/error-handling)
[Introduction](https://docs.runanywhere.ai/flutter/introduction)

⌘I

---

# Error Handling - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/error-handling#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

React Native SDK

Error Handling

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/error-handling#overview)
    
*   [SDKError Structure](https://docs.runanywhere.ai/react-native/error-handling#sdkerror-structure)
    
*   [Checking for SDK Errors](https://docs.runanywhere.ai/react-native/error-handling#checking-for-sdk-errors)
    
*   [Error Codes](https://docs.runanywhere.ai/react-native/error-handling#error-codes)
    
*   [General Errors](https://docs.runanywhere.ai/react-native/error-handling#general-errors)
    
*   [Model Errors](https://docs.runanywhere.ai/react-native/error-handling#model-errors)
    
*   [Generation Errors](https://docs.runanywhere.ai/react-native/error-handling#generation-errors)
    
*   [Voice Errors](https://docs.runanywhere.ai/react-native/error-handling#voice-errors)
    
*   [Resource Errors](https://docs.runanywhere.ai/react-native/error-handling#resource-errors)
    
*   [Network/Auth Errors](https://docs.runanywhere.ai/react-native/error-handling#network%2Fauth-errors)
    
*   [Error Categories](https://docs.runanywhere.ai/react-native/error-handling#error-categories)
    
*   [Handling Patterns](https://docs.runanywhere.ai/react-native/error-handling#handling-patterns)
    
*   [Comprehensive Error Handler](https://docs.runanywhere.ai/react-native/error-handling#comprehensive-error-handler)
    
*   [Category-Based Handling](https://docs.runanywhere.ai/react-native/error-handling#category-based-handling)
    
*   [React Hook for Error Handling](https://docs.runanywhere.ai/react-native/error-handling#react-hook-for-error-handling)
    
*   [Error Display Component](https://docs.runanywhere.ai/react-native/error-handling#error-display-component)
    
*   [Creating Custom Errors](https://docs.runanywhere.ai/react-native/error-handling#creating-custom-errors)
    
*   [Logging Errors](https://docs.runanywhere.ai/react-native/error-handling#logging-errors)
    
*   [Related](https://docs.runanywhere.ai/react-native/error-handling#related)
    

[​](https://docs.runanywhere.ai/react-native/error-handling#overview)

Overview
---------------------------------------------------------------------------------

The RunAnywhere SDK provides structured error handling through `SDKError` with specific error codes and categories. This guide covers error types, handling patterns, and recovery strategies.

[​](https://docs.runanywhere.ai/react-native/error-handling#sdkerror-structure)

SDKError Structure
-----------------------------------------------------------------------------------------------------

Copy

Ask AI

    interface SDKError extends Error {
      /** Unique error code */
      code: SDKErrorCode
    
      /** Error category */
      category: ErrorCategory
    
      /** Original error (if wrapped) */
      underlyingError?: Error
    
      /** Additional context */
      context?: ErrorContext
    
      /** Suggested recovery action */
      recoverySuggestion?: string
    }
    

[​](https://docs.runanywhere.ai/react-native/error-handling#checking-for-sdk-errors)

Checking for SDK Errors
---------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    import { isSDKError, SDKError, SDKErrorCode } from '@runanywhere/core'
    
    try {
      await RunAnywhere.generate('Hello')
    } catch (error) {
      if (isSDKError(error)) {
        // Handle SDK-specific error
        console.log('Code:', error.code)
        console.log('Category:', error.category)
        console.log('Message:', error.message)
        console.log('Recovery:', error.recoverySuggestion)
      } else {
        // Handle generic error
        console.error('Unexpected error:', error)
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/error-handling#error-codes)

Error Codes
---------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/error-handling#general-errors)

General Errors

| Code | Description | Recovery |
| --- | --- | --- |
| `notInitialized` | SDK not initialized | Call `RunAnywhere.initialize()` |
| `alreadyInitialized` | SDK already initialized | Use existing instance or call `reset()` |
| `invalidInput` | Invalid input parameters | Check input values |
| `nativeModuleNotAvailable` | Native module not linked | Rebuild app, check pod install |

### 

[​](https://docs.runanywhere.ai/react-native/error-handling#model-errors)

Model Errors

| Code | Description | Recovery |
| --- | --- | --- |
| `modelNotFound` | Model not in registry | Add model with `LlamaCPP.addModel()` |
| `modelLoadFailed` | Failed to load model | Check model path, format, memory |
| `modelNotLoaded` | No model loaded | Call `loadModel()` first |
| `modelDownloadFailed` | Download failed | Check network, retry |

### 

[​](https://docs.runanywhere.ai/react-native/error-handling#generation-errors)

Generation Errors

| Code | Description | Recovery |
| --- | --- | --- |
| `generationFailed` | Text generation failed | Check model, reduce tokens |
| `generationCancelled` | Generation was cancelled | Expected if `cancelGeneration()` called |

### 

[​](https://docs.runanywhere.ai/react-native/error-handling#voice-errors)

Voice Errors

| Code | Description | Recovery |
| --- | --- | --- |
| `sttFailed` | Transcription failed | Check audio format, model |
| `ttsFailed` | Synthesis failed | Check TTS model, text |
| `vadFailed` | VAD failed | Check VAD configuration |
| `voiceAgentFailed` | Voice pipeline failed | Check all models loaded |

### 

[​](https://docs.runanywhere.ai/react-native/error-handling#resource-errors)

Resource Errors

| Code | Description | Recovery |
| --- | --- | --- |
| `insufficientStorage` | Not enough disk space | Free storage, delete models |
| `insufficientMemory` | Not enough RAM | Unload models, use smaller model |
| `downloadFailed` | Download failed | Check network, storage |
| `downloadCancelled` | Download cancelled | Expected if cancelled by user |

### 

[​](https://docs.runanywhere.ai/react-native/error-handling#network/auth-errors)

Network/Auth Errors

| Code | Description | Recovery |
| --- | --- | --- |
| `networkUnavailable` | No network connection | Check connectivity |
| `authenticationFailed` | Invalid API key | Check API key |
| `permissionDenied` | Missing permission | Request permission |

[​](https://docs.runanywhere.ai/react-native/error-handling#error-categories)

Error Categories
-------------------------------------------------------------------------------------------------

Copy

Ask AI

    enum ErrorCategory {
      General = 'general',
      LLM = 'llm',
      STT = 'stt',
      TTS = 'tts',
      VAD = 'vad',
      VoiceAgent = 'voiceAgent',
      Download = 'download',
      Network = 'network',
      Authentication = 'authentication',
      Storage = 'storage',
      Memory = 'memory',
    }
    

[​](https://docs.runanywhere.ai/react-native/error-handling#handling-patterns)

Handling Patterns
---------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/error-handling#comprehensive-error-handler)

Comprehensive Error Handler

Copy

Ask AI

    import { isSDKError, SDKError, SDKErrorCode, ErrorCategory } from '@runanywhere/core'
    
    async function handleSDKOperation<T>(
      operation: () => Promise<T>,
      options?: {
        onRetry?: () => void
        maxRetries?: number
      }
    ): Promise<T> {
      const maxRetries = options?.maxRetries ?? 1
      let lastError: Error | null = null
    
      for (let attempt = 0; attempt <= maxRetries; attempt++) {
        try {
          return await operation()
        } catch (error) {
          lastError = error as Error
    
          if (!isSDKError(error)) {
            throw error // Non-SDK error, don't retry
          }
    
          // Handle specific errors
          switch (error.code) {
            // Retryable errors
            case SDKErrorCode.networkUnavailable:
            case SDKErrorCode.downloadFailed:
              if (attempt < maxRetries) {
                options?.onRetry?.()
                await delay(1000 * (attempt + 1)) // Exponential backoff
                continue
              }
              break
    
            // Non-retryable errors
            case SDKErrorCode.notInitialized:
            case SDKErrorCode.modelNotFound:
            case SDKErrorCode.invalidInput:
            case SDKErrorCode.insufficientStorage:
              throw error
    
            // Memory errors - try recovery
            case SDKErrorCode.insufficientMemory:
              await RunAnywhere.unloadModel()
              if (attempt < maxRetries) {
                continue
              }
              break
          }
    
          throw error
        }
      }
    
      throw lastError
    }
    
    function delay(ms: number): Promise<void> {
      return new Promise((resolve) => setTimeout(resolve, ms))
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/error-handling#category-based-handling)

Category-Based Handling

Copy

Ask AI

    function handleError(error: SDKError): string {
      switch (error.category) {
        case ErrorCategory.LLM:
          return handleLLMError(error)
        case ErrorCategory.STT:
          return handleSTTError(error)
        case ErrorCategory.TTS:
          return handleTTSError(error)
        case ErrorCategory.Network:
          return 'Please check your internet connection.'
        case ErrorCategory.Storage:
          return 'Not enough storage space. Please free up some space.'
        case ErrorCategory.Memory:
          return 'Not enough memory. Try closing other apps.'
        default:
          return error.recoverySuggestion || error.message
      }
    }
    
    function handleLLMError(error: SDKError): string {
      switch (error.code) {
        case SDKErrorCode.modelNotLoaded:
          return 'Please wait for the AI model to load.'
        case SDKErrorCode.generationFailed:
          return 'Failed to generate response. Please try again.'
        case SDKErrorCode.generationCancelled:
          return 'Generation was cancelled.'
        default:
          return 'An error occurred with the AI model.'
      }
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/error-handling#react-hook-for-error-handling)

React Hook for Error Handling

useSDKError.ts

Copy

Ask AI

    import { useState, useCallback } from 'react'
    import { isSDKError, SDKError, SDKErrorCode } from '@runanywhere/core'
    
    interface ErrorState {
      error: SDKError | Error | null
      userMessage: string | null
      canRetry: boolean
    }
    
    export function useSDKError() {
      const [errorState, setErrorState] = useState<ErrorState>({
        error: null,
        userMessage: null,
        canRetry: false,
      })
    
      const handleError = useCallback((error: unknown) => {
        if (isSDKError(error)) {
          const canRetry = [\
            SDKErrorCode.networkUnavailable,\
            SDKErrorCode.downloadFailed,\
            SDKErrorCode.generationFailed,\
          ].includes(error.code)
    
          setErrorState({
            error,
            userMessage: getErrorMessage(error),
            canRetry,
          })
        } else {
          setErrorState({
            error: error as Error,
            userMessage: 'An unexpected error occurred.',
            canRetry: false,
          })
        }
      }, [])
    
      const clearError = useCallback(() => {
        setErrorState({ error: null, userMessage: null, canRetry: false })
      }, [])
    
      return { ...errorState, handleError, clearError }
    }
    
    function getErrorMessage(error: SDKError): string {
      // Return user-friendly messages based on error code
      switch (error.code) {
        case SDKErrorCode.notInitialized:
          return 'The app is still loading. Please wait.'
        case SDKErrorCode.modelNotLoaded:
          return 'The AI model is not ready yet.'
        case SDKErrorCode.networkUnavailable:
          return 'No internet connection.'
        case SDKErrorCode.insufficientMemory:
          return 'Not enough memory. Try closing other apps.'
        case SDKErrorCode.insufficientStorage:
          return 'Not enough storage space.'
        default:
          return error.recoverySuggestion || error.message
      }
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/error-handling#error-display-component)

Error Display Component

ErrorDisplay.tsx

Copy

Ask AI

    import React from 'react'
    import { View, Text, Button, StyleSheet } from 'react-native'
    import { SDKError } from '@runanywhere/core'
    
    interface Props {
      error: SDKError | Error
      message: string
      canRetry: boolean
      onRetry?: () => void
      onDismiss?: () => void
    }
    
    export function ErrorDisplay({ error, message, canRetry, onRetry, onDismiss }: Props) {
      return (
        <View style={styles.container}>
          <Text style={styles.icon}>⚠️</Text>
          <Text style={styles.message}>{message}</Text>
    
          <View style={styles.buttons}>
            {canRetry && onRetry && <Button title="Retry" onPress={onRetry} />}
            {onDismiss && <Button title="Dismiss" onPress={onDismiss} color="#666" />}
          </View>
        </View>
      )
    }
    
    const styles = StyleSheet.create({
      container: {
        padding: 16,
        backgroundColor: '#fff3cd',
        borderRadius: 8,
        alignItems: 'center',
      },
      icon: { fontSize: 32, marginBottom: 8 },
      message: { fontSize: 16, textAlign: 'center', marginBottom: 16 },
      buttons: { flexDirection: 'row', gap: 8 },
    })
    

[​](https://docs.runanywhere.ai/react-native/error-handling#creating-custom-errors)

Creating Custom Errors
-------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    import {
      notInitializedError,
      modelNotFoundError,
      modelLoadError,
      generationError,
      networkError,
    } from '@runanywhere/core'
    
    // Use factory functions
    throw notInitializedError()
    throw modelNotFoundError('my-model-id')
    throw generationError('Context window exceeded')
    

[​](https://docs.runanywhere.ai/react-native/error-handling#logging-errors)

Logging Errors
---------------------------------------------------------------------------------------------

Copy

Ask AI

    import { SDKLogger } from '@runanywhere/core'
    
    const logger = new SDKLogger('MyApp')
    
    try {
      await RunAnywhere.generate(prompt)
    } catch (error) {
      if (isSDKError(error)) {
        logger.error('SDK Error', error, {
          code: error.code,
          category: error.category,
        })
      } else {
        logger.error('Unexpected error', error as Error)
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/error-handling#related)

Related
-------------------------------------------------------------------------------

[Configuration\
-------------\
\
SDK configuration](https://docs.runanywhere.ai/react-native/configuration)
[Best Practices\
--------------\
\
Optimization tips](https://docs.runanywhere.ai/react-native/best-practices)

[Configuration](https://docs.runanywhere.ai/react-native/configuration)
[Best Practices](https://docs.runanywhere.ai/react-native/best-practices)

⌘I

---

# STT Options - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/stt/options#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

STT Options

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/stt/options#overview)
    
*   [Options Reference](https://docs.runanywhere.ai/react-native/stt/options#options-reference)
    
*   [Language Support](https://docs.runanywhere.ai/react-native/stt/options#language-support)
    
*   [Supported Languages](https://docs.runanywhere.ai/react-native/stt/options#supported-languages)
    
*   [Punctuation](https://docs.runanywhere.ai/react-native/stt/options#punctuation)
    
*   [Word Timestamps](https://docs.runanywhere.ai/react-native/stt/options#word-timestamps)
    
*   [Use Cases](https://docs.runanywhere.ai/react-native/stt/options#use-cases)
    
*   [Example: Subtitle Generator](https://docs.runanywhere.ai/react-native/stt/options#example-subtitle-generator)
    
*   [Speaker Diarization](https://docs.runanywhere.ai/react-native/stt/options#speaker-diarization)
    
*   [Sample Rate](https://docs.runanywhere.ai/react-native/stt/options#sample-rate)
    
*   [Model Loading Options](https://docs.runanywhere.ai/react-native/stt/options#model-loading-options)
    
*   [Combining Options](https://docs.runanywhere.ai/react-native/stt/options#combining-options)
    
*   [Performance vs Accuracy](https://docs.runanywhere.ai/react-native/stt/options#performance-vs-accuracy)
    
*   [Related](https://docs.runanywhere.ai/react-native/stt/options#related)
    

[​](https://docs.runanywhere.ai/react-native/stt/options#overview)

Overview
------------------------------------------------------------------------------

The STT API provides various options to customize transcription behavior, from language selection to word-level timestamps.

[​](https://docs.runanywhere.ai/react-native/stt/options#options-reference)

Options Reference
------------------------------------------------------------------------------------------------

Copy

Ask AI

    interface STTOptions {
      /** Language code for transcription */
      language?: string
    
      /** Enable punctuation in output */
      punctuation?: boolean
    
      /** Enable speaker diarization (multi-speaker) */
      diarization?: boolean
    
      /** Enable word-level timestamps */
      wordTimestamps?: boolean
    
      /** Audio sample rate (default: 16000) */
      sampleRate?: number
    }
    

[​](https://docs.runanywhere.ai/react-native/stt/options#language-support)

Language Support
----------------------------------------------------------------------------------------------

Specify the language code to improve accuracy:

Copy

Ask AI

    // English
    const english = await RunAnywhere.transcribe(audioBase64, { language: 'en' })
    
    // Spanish
    const spanish = await RunAnywhere.transcribe(audioBase64, { language: 'es' })
    
    // French
    const french = await RunAnywhere.transcribe(audioBase64, { language: 'fr' })
    
    // Auto-detect (slower)
    const auto = await RunAnywhere.transcribe(audioBase64)
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/options#supported-languages)

Supported Languages

| Code | Language | Code | Language |
| --- | --- | --- | --- |
| `en` | English | `ja` | Japanese |
| `es` | Spanish | `ko` | Korean |
| `fr` | French | `pt` | Portuguese |
| `de` | German | `ru` | Russian |
| `it` | Italian | `zh` | Chinese |
| `nl` | Dutch | `ar` | Arabic |
| `pl` | Polish | `hi` | Hindi |

Language-specific models (e.g., `whisper-tiny.en`) only support that language but are more accurate and faster.

[​](https://docs.runanywhere.ai/react-native/stt/options#punctuation)

Punctuation
------------------------------------------------------------------------------------

Add punctuation to transcription output:

Copy

Ask AI

    // Without punctuation (default for some models)
    const noPunct = await RunAnywhere.transcribe(audioBase64, {
      language: 'en',
      punctuation: false,
    })
    // "hello how are you today"
    
    // With punctuation
    const withPunct = await RunAnywhere.transcribe(audioBase64, {
      language: 'en',
      punctuation: true,
    })
    // "Hello, how are you today?"
    

[​](https://docs.runanywhere.ai/react-native/stt/options#word-timestamps)

Word Timestamps
--------------------------------------------------------------------------------------------

Get timing information for each word:

Copy

Ask AI

    const result = await RunAnywhere.transcribe(audioBase64, {
      language: 'en',
      wordTimestamps: true,
    })
    
    console.log('Transcription:', result.text)
    
    // Each segment contains word-level timing
    for (const segment of result.segments) {
      console.log(`[${segment.startTime.toFixed(2)}s - ${segment.endTime.toFixed(2)}s] ${segment.text}`)
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/options#use-cases)

Use Cases

*   **Subtitles/Captions**: Sync text with video
*   **Karaoke**: Highlight words as they’re spoken
*   **Search**: Jump to specific moments in audio
*   **Accessibility**: Show words as they’re spoken

### 

[​](https://docs.runanywhere.ai/react-native/stt/options#example-subtitle-generator)

Example: Subtitle Generator

Copy

Ask AI

    interface Subtitle {
      start: number
      end: number
      text: string
    }
    
    async function generateSubtitles(audioBase64: string): Promise<Subtitle[]> {
      const result = await RunAnywhere.transcribe(audioBase64, {
        language: 'en',
        wordTimestamps: true,
      })
    
      return result.segments.map((segment) => ({
        start: segment.startTime,
        end: segment.endTime,
        text: segment.text.trim(),
      }))
    }
    
    // Convert to SRT format
    function toSRT(subtitles: Subtitle[]): string {
      return subtitles
        .map((sub, i) => {
          const start = formatSRTTime(sub.start)
          const end = formatSRTTime(sub.end)
          return `${i + 1}\n${start} --> ${end}\n${sub.text}\n`
        })
        .join('\n')
    }
    
    function formatSRTTime(seconds: number): string {
      const h = Math.floor(seconds / 3600)
      const m = Math.floor((seconds % 3600) / 60)
      const s = Math.floor(seconds % 60)
      const ms = Math.floor((seconds % 1) * 1000)
      return `${h.toString().padStart(2, '0')}:${m.toString().padStart(2, '0')}:${s.toString().padStart(2, '0')},${ms.toString().padStart(3, '0')}`
    }
    

[​](https://docs.runanywhere.ai/react-native/stt/options#speaker-diarization)

Speaker Diarization
----------------------------------------------------------------------------------------------------

Identify different speakers in the audio:

Copy

Ask AI

    const result = await RunAnywhere.transcribe(audioBase64, {
      language: 'en',
      diarization: true,
    })
    
    // Segments include speaker IDs
    for (const segment of result.segments) {
      console.log(`[Speaker ${segment.speakerId}]: ${segment.text}`)
    }
    

Speaker diarization is computationally expensive and may not be available on all models. Check model documentation for support.

[​](https://docs.runanywhere.ai/react-native/stt/options#sample-rate)

Sample Rate
------------------------------------------------------------------------------------

Specify the audio sample rate if different from the default:

Copy

Ask AI

    // Standard 16kHz (default, recommended for STT)
    const standard = await RunAnywhere.transcribeBuffer(samples)
    
    // Higher quality 44.1kHz (will be downsampled internally)
    const highQuality = await RunAnywhere.transcribeBuffer(samples, {
      sampleRate: 44100,
    })
    

`transcribeBuffer()` accepts a `Float32Array` of PCM audio samples. The sample rate defaults to 16000 Hz if not specified in options.

For best results, record audio at 16kHz mono. Higher sample rates will be downsampled, which adds processing overhead.

[​](https://docs.runanywhere.ai/react-native/stt/options#model-loading-options)

Model Loading Options
--------------------------------------------------------------------------------------------------------

Configure model loading:

Copy

Ask AI

    // Load STT model with specific type
    await RunAnywhere.loadSTTModel(modelPath, 'whisper')
    
    // Check if model is loaded
    const isLoaded = await RunAnywhere.isSTTModelLoaded()
    
    // Unload when done
    await RunAnywhere.unloadSTTModel()
    

[​](https://docs.runanywhere.ai/react-native/stt/options#combining-options)

Combining Options
------------------------------------------------------------------------------------------------

Copy

Ask AI

    // Full-featured transcription
    const result = await RunAnywhere.transcribe(audioBase64, {
      language: 'en',
      punctuation: true,
      wordTimestamps: true,
    })
    
    // Access all result data
    console.log('Text:', result.text)
    console.log('Language:', result.language)
    console.log('Confidence:', result.confidence)
    console.log('Duration:', result.duration, 'seconds')
    console.log('Segments:', result.segments.length)
    
    // Alternatives (if available)
    if (result.alternatives.length > 0) {
      console.log('Alternatives:')
      result.alternatives.forEach((alt, i) => {
        console.log(`  ${i + 1}. ${alt.text} (confidence: ${alt.confidence})`)
      })
    }
    

[​](https://docs.runanywhere.ai/react-native/stt/options#performance-vs-accuracy)

Performance vs Accuracy
------------------------------------------------------------------------------------------------------------

| Option | Impact on Speed | Impact on Accuracy |
| --- | --- | --- |
| `language` specified | Faster | Better |
| `wordTimestamps` | Slower | Same |
| `diarization` | Much slower | Same |
| `punctuation` | Minimal | Same |

For best performance, always specify the `language` option rather than relying on auto-detection.

[​](https://docs.runanywhere.ai/react-native/stt/options#related)

Related
----------------------------------------------------------------------------

[Transcribe\
----------\
\
Basic transcription](https://docs.runanywhere.ai/react-native/stt/transcribe)
[STT Streaming\
-------------\
\
Real-time transcription](https://docs.runanywhere.ai/react-native/stt/stream)
[VAD\
---\
\
Voice Activity Detection](https://docs.runanywhere.ai/react-native/vad)

[Streaming STT](https://docs.runanywhere.ai/react-native/stt/stream)
[synthesize()](https://docs.runanywhere.ai/react-native/tts/synthesize)

⌘I

---

# Error Handling - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/error-handling#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Swift SDK

Error Handling

On this page

*   [SDKError](https://docs.runanywhere.ai/swift/error-handling#sdkerror)
    
*   [Error Codes](https://docs.runanywhere.ai/swift/error-handling#error-codes)
    
*   [Error Categories](https://docs.runanywhere.ai/swift/error-handling#error-categories)
    
*   [Basic Error Handling](https://docs.runanywhere.ai/swift/error-handling#basic-error-handling)
    
*   [Handling Specific Errors](https://docs.runanywhere.ai/swift/error-handling#handling-specific-errors)
    
*   [By Error Code](https://docs.runanywhere.ai/swift/error-handling#by-error-code)
    
*   [By Category](https://docs.runanywhere.ai/swift/error-handling#by-category)
    
*   [Error Factory Methods](https://docs.runanywhere.ai/swift/error-handling#error-factory-methods)
    
*   [SwiftUI Error Handling](https://docs.runanywhere.ai/swift/error-handling#swiftui-error-handling)
    
*   [Error Recovery Strategies](https://docs.runanywhere.ai/swift/error-handling#error-recovery-strategies)
    
*   [Retry Logic](https://docs.runanywhere.ai/swift/error-handling#retry-logic)
    
*   [Graceful Degradation](https://docs.runanywhere.ai/swift/error-handling#graceful-degradation)
    
*   [Model Fallback](https://docs.runanywhere.ai/swift/error-handling#model-fallback)
    
*   [Logging Errors](https://docs.runanywhere.ai/swift/error-handling#logging-errors)
    
*   [Event-Based Error Handling](https://docs.runanywhere.ai/swift/error-handling#event-based-error-handling)
    
*   [Best Practices](https://docs.runanywhere.ai/swift/error-handling#best-practices)
    

The RunAnywhere SDK uses a unified error system that provides detailed information for debugging and user-friendly messages for display.

[​](https://docs.runanywhere.ai/swift/error-handling#sdkerror)

SDKError
--------------------------------------------------------------------------

All SDK errors are represented by `SDKError`:

Copy

Ask AI

    public struct SDKError: Error, LocalizedError, Sendable {
        public let code: ErrorCode           // Error code enum
        public let message: String           // Detailed message
        public let category: ErrorCategory   // Error category
        public let stackTrace: [String]      // Debug stack trace
        public let underlyingError: (any Error)? // Wrapped error
    
        public var errorDescription: String? { get }
        public var failureReason: String? { get }
        public var recoverySuggestion: String? { get }
    }
    

[​](https://docs.runanywhere.ai/swift/error-handling#error-codes)

Error Codes
--------------------------------------------------------------------------------

Copy

Ask AI

    public enum ErrorCode: String, Sendable {
        // Initialization
        case notInitialized
        case invalidConfiguration
        case invalidAPIKey
        case validationFailed
    
        // Models
        case modelNotFound
        case modelLoadFailed
        case modelIncompatible
    
        // Generation
        case generationFailed
        case processingFailed
        case streamingNotSupported
    
        // Audio
        case emptyAudioBuffer
        case microphonePermissionDenied
    
        // Network
        case networkUnavailable
        case networkError
        case timeout
    
        // Storage
        case insufficientStorage
        case insufficientMemory
        case storageError
    
        // General
        case initializationFailed
        case cancelled
        case invalidState
        case invalidInput
        case notImplemented
        case unknown
    }
    

[​](https://docs.runanywhere.ai/swift/error-handling#error-categories)

Error Categories
------------------------------------------------------------------------------------------

Copy

Ask AI

    public enum ErrorCategory: String, Sendable {
        case general
        case stt
        case tts
        case llm
        case vad
        case voiceAgent
        case download
        case fileManagement
        case network
        case authentication
        case security
        case runtime
    }
    

[​](https://docs.runanywhere.ai/swift/error-handling#basic-error-handling)

Basic Error Handling
--------------------------------------------------------------------------------------------------

Copy

Ask AI

    do {
        let result = try await RunAnywhere.generate("Hello")
        print(result.text)
    } catch let error as SDKError {
        print("Error: \(error.message)")
        print("Code: \(error.code)")
        print("Category: \(error.category)")
    
        if let suggestion = error.recoverySuggestion {
            print("Suggestion: \(suggestion)")
        }
    } catch {
        print("Unexpected error: \(error)")
    }
    

[​](https://docs.runanywhere.ai/swift/error-handling#handling-specific-errors)

Handling Specific Errors
----------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/error-handling#by-error-code)

By Error Code

Copy

Ask AI

    do {
        let result = try await RunAnywhere.generate(prompt)
    } catch let error as SDKError {
        switch error.code {
        case .notInitialized:
            // SDK not initialized
            showAlert("Please restart the app")
    
        case .modelNotFound:
            // Model not loaded
            try await RunAnywhere.loadModel("default-model")
    
        case .modelLoadFailed:
            // Model failed to load
            showAlert("Failed to load AI model. Try redownloading.")
    
        case .generationFailed:
            // Generation failed
            showAlert("Generation failed. Please try again.")
    
        case .insufficientMemory:
            // Not enough RAM
            try await RunAnywhere.unloadModel()
            showAlert("Low memory. Some features may be limited.")
    
        case .networkUnavailable:
            // No network (may be needed for auth)
            showOfflineMode()
    
        case .timeout:
            // Operation timed out
            showAlert("Request timed out. Please try again.")
    
        case .cancelled:
            // User cancelled
            break
    
        default:
            showAlert("An error occurred: \(error.message)")
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/error-handling#by-category)

By Category

Copy

Ask AI

    do {
        let result = try await RunAnywhere.processVoiceTurn(audioData)
    } catch let error as SDKError {
        switch error.category {
        case .llm:
            handleLLMError(error)
        case .stt:
            handleSTTError(error)
        case .tts:
            handleTTSError(error)
        case .network:
            handleNetworkError(error)
        case .authentication:
            handleAuthError(error)
        default:
            handleGenericError(error)
        }
    }
    

[​](https://docs.runanywhere.ai/swift/error-handling#error-factory-methods)

Error Factory Methods
----------------------------------------------------------------------------------------------------

The SDK provides factory methods for creating errors:

Copy

Ask AI

    // Create errors programmatically
    SDKError.general(.notInitialized, "SDK not initialized")
    SDKError.llm(.generationFailed, "Generation failed", underlying: originalError)
    SDKError.stt(.modelNotFound, "STT model not found")
    SDKError.tts(.processingFailed, "TTS synthesis failed")
    SDKError.network(.networkUnavailable, "No internet connection")
    SDKError.authentication(.invalidAPIKey, "Invalid API key")
    

[​](https://docs.runanywhere.ai/swift/error-handling#swiftui-error-handling)

SwiftUI Error Handling
------------------------------------------------------------------------------------------------------

Copy

Ask AI

    struct ContentView: View {
        @State private var error: SDKError?
        @State private var showError = false
        @State private var response = ""
    
        var body: some View {
            VStack {
                Text(response)
    
                Button("Generate") {
                    Task { await generate() }
                }
            }
            .alert("Error", isPresented: $showError, presenting: error) { _ in
                Button("OK") { error = nil }
            } message: { error in
                VStack {
                    Text(error.message)
                    if let suggestion = error.recoverySuggestion {
                        Text(suggestion)
                            .font(.caption)
                    }
                }
            }
        }
    
        func generate() async {
            do {
                let result = try await RunAnywhere.generate("Hello")
                response = result.text
            } catch let sdkError as SDKError {
                error = sdkError
                showError = true
            } catch {
                self.error = SDKError.general(.unknown, error.localizedDescription)
                showError = true
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/error-handling#error-recovery-strategies)

Error Recovery Strategies
------------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/error-handling#retry-logic)

Retry Logic

Copy

Ask AI

    func generateWithRetry(
        prompt: String,
        maxRetries: Int = 3
    ) async throws -> LLMGenerationResult {
        var lastError: Error?
    
        for attempt in 1...maxRetries {
            do {
                return try await RunAnywhere.generate(prompt)
            } catch let error as SDKError {
                lastError = error
    
                // Don't retry certain errors
                switch error.code {
                case .notInitialized, .invalidAPIKey, .modelNotFound:
                    throw error
                case .timeout, .networkError:
                    // Retry with delay
                    try await Task.sleep(for: .seconds(Double(attempt)))
                default:
                    throw error
                }
            }
        }
    
        throw lastError ?? SDKError.general(.unknown, "Max retries exceeded")
    }
    

### 

[​](https://docs.runanywhere.ai/swift/error-handling#graceful-degradation)

Graceful Degradation

Copy

Ask AI

    func generateResponse(_ prompt: String) async -> String {
        do {
            let result = try await RunAnywhere.generate(prompt)
            return result.text
        } catch let error as SDKError {
            switch error.code {
            case .modelNotFound:
                return "Please wait while the AI model loads..."
            case .insufficientMemory:
                return "Memory is low. Please close other apps and try again."
            case .timeout:
                return "The request took too long. Please try a shorter prompt."
            default:
                return "I'm having trouble responding right now. Please try again."
            }
        } catch {
            return "An unexpected error occurred."
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/error-handling#model-fallback)

Model Fallback

Copy

Ask AI

    func generateWithFallback(_ prompt: String) async throws -> String {
        // Try primary model
        do {
            return try await RunAnywhere.chat(prompt)
        } catch let error as SDKError where error.code == .generationFailed {
            // Try smaller model
            try await RunAnywhere.unloadModel()
            try await RunAnywhere.loadModel("llama-3.2-1b-instruct-q4")  // Smaller model
            return try await RunAnywhere.chat(prompt)
        }
    }
    

[​](https://docs.runanywhere.ai/swift/error-handling#logging-errors)

Logging Errors
--------------------------------------------------------------------------------------

Copy

Ask AI

    func logError(_ error: SDKError) {
        print("""
        === SDK Error ===
        Code: \(error.code.rawValue)
        Category: \(error.category.rawValue)
        Message: \(error.message)
        Suggestion: \(error.recoverySuggestion ?? "None")
        Stack: \(error.stackTrace.joined(separator: "\n"))
        Underlying: \(error.underlyingError?.localizedDescription ?? "None")
        ================
        """)
    }
    

[​](https://docs.runanywhere.ai/swift/error-handling#event-based-error-handling)

Event-Based Error Handling
--------------------------------------------------------------------------------------------------------------

Subscribe to error events SDK-wide:

Copy

Ask AI

    import Combine
    
    class ErrorHandler {
        private var cancellables = Set<AnyCancellable>()
    
        init() {
            // Subscribe to all errors
            RunAnywhere.events.events
                .compactMap { $0 as? ErrorEvent }
                .sink { event in
                    self.handleError(event.error)
                }
                .store(in: &cancellables)
        }
    
        private func handleError(_ error: Error) {
            // Log to analytics, show UI, etc.
        }
    }
    

[​](https://docs.runanywhere.ai/swift/error-handling#best-practices)

Best Practices
--------------------------------------------------------------------------------------

Always catch SDKError

Catch `SDKError` specifically to access detailed error information.

Provide user-friendly messages

Use `recoverySuggestion` when available, or map error codes to friendly messages.

Log errors for debugging

Log full error details including stack traces in development builds.

Implement recovery

Where possible, implement automatic recovery (retries, fallbacks).

[Configuration\
-------------\
\
SDK configuration →](https://docs.runanywhere.ai/swift/configuration)
[Best Practices\
--------------\
\
Performance tips →](https://docs.runanywhere.ai/swift/best-practices)

[Configuration](https://docs.runanywhere.ai/swift/configuration)
[Best Practices](https://docs.runanywhere.ai/swift/best-practices)

⌘I

---

# Best Practices - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/best-practices#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Swift SDK

Best Practices

On this page

*   [Initialization](https://docs.runanywhere.ai/swift/best-practices#initialization)
    
*   [Initialize Early](https://docs.runanywhere.ai/swift/best-practices#initialize-early)
    
*   [Pre-load Models](https://docs.runanywhere.ai/swift/best-practices#pre-load-models)
    
*   [Memory Management](https://docs.runanywhere.ai/swift/best-practices#memory-management)
    
*   [Unload When Not Needed](https://docs.runanywhere.ai/swift/best-practices#unload-when-not-needed)
    
*   [Monitor Memory](https://docs.runanywhere.ai/swift/best-practices#monitor-memory)
    
*   [Handle Memory Warnings](https://docs.runanywhere.ai/swift/best-practices#handle-memory-warnings)
    
*   [Model Selection](https://docs.runanywhere.ai/swift/best-practices#model-selection)
    
*   [Choose Appropriate Model Sizes](https://docs.runanywhere.ai/swift/best-practices#choose-appropriate-model-sizes)
    
*   [Device-Specific Loading](https://docs.runanywhere.ai/swift/best-practices#device-specific-loading)
    
*   [Streaming for Responsiveness](https://docs.runanywhere.ai/swift/best-practices#streaming-for-responsiveness)
    
*   [Always Stream for Long Outputs](https://docs.runanywhere.ai/swift/best-practices#always-stream-for-long-outputs)
    
*   [Batch UI Updates](https://docs.runanywhere.ai/swift/best-practices#batch-ui-updates)
    
*   [Threading](https://docs.runanywhere.ai/swift/best-practices#threading)
    
*   [Use Appropriate Actors](https://docs.runanywhere.ai/swift/best-practices#use-appropriate-actors)
    
*   [Don’t Block the Main Thread](https://docs.runanywhere.ai/swift/best-practices#don%E2%80%99t-block-the-main-thread)
    
*   [Error Recovery](https://docs.runanywhere.ai/swift/best-practices#error-recovery)
    
*   [Implement Retries](https://docs.runanywhere.ai/swift/best-practices#implement-retries)
    
*   [Graceful Fallbacks](https://docs.runanywhere.ai/swift/best-practices#graceful-fallbacks)
    
*   [Storage](https://docs.runanywhere.ai/swift/best-practices#storage)
    
*   [Check Before Downloads](https://docs.runanywhere.ai/swift/best-practices#check-before-downloads)
    
*   [Clean Up Regularly](https://docs.runanywhere.ai/swift/best-practices#clean-up-regularly)
    
*   [Event Handling](https://docs.runanywhere.ai/swift/best-practices#event-handling)
    
*   [Subscribe to Events](https://docs.runanywhere.ai/swift/best-practices#subscribe-to-events)
    
*   [Testing](https://docs.runanywhere.ai/swift/best-practices#testing)
    
*   [Use Development Environment](https://docs.runanywhere.ai/swift/best-practices#use-development-environment)
    
*   [Mock for Unit Tests](https://docs.runanywhere.ai/swift/best-practices#mock-for-unit-tests)
    
*   [Quick Reference](https://docs.runanywhere.ai/swift/best-practices#quick-reference)
    

Follow these best practices to build high-performance, reliable apps with the RunAnywhere SDK.

[​](https://docs.runanywhere.ai/swift/best-practices#initialization)

Initialization
--------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/best-practices#initialize-early)

Initialize Early

Initialize the SDK at app launch, not when first needed:

Copy

Ask AI

    @main
    struct MyApp: App {
        init() {
            Task { @MainActor in
                // Register modules
                LlamaCPP.register()
                ONNX.register()
    
                // Initialize SDK
                try? RunAnywhere.initialize(
                    apiKey: Config.apiKey,
                    baseURL: Config.baseURL,
                    environment: .production
                )
            }
        }
    
        var body: some Scene {
            WindowGroup {
                ContentView()
            }
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/best-practices#pre-load-models)

Pre-load Models

Load models during onboarding or splash screen:

Copy

Ask AI

    struct SplashView: View {
        @State private var loadingProgress = 0.0
        @State private var isReady = false
    
        var body: some View {
            VStack {
                ProgressView(value: loadingProgress)
                Text("Loading AI models...")
            }
            .task {
                await loadModels()
            }
        }
    
        func loadModels() async {
            // Load LLM (50%)
            try? await RunAnywhere.loadModel("llama-3.2-1b-instruct-q4")
            loadingProgress = 0.5
    
            // Load STT (75%)
            try? await RunAnywhere.loadSTTModel("whisper-base-onnx")
            loadingProgress = 0.75
    
            // Load TTS (100%)
            try? await RunAnywhere.loadTTSVoice("piper-en-us-amy")
            loadingProgress = 1.0
    
            isReady = true
        }
    }
    

[​](https://docs.runanywhere.ai/swift/best-practices#memory-management)

Memory Management
--------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/best-practices#unload-when-not-needed)

Unload When Not Needed

Free memory by unloading unused models:

Copy

Ask AI

    class ModelManager {
        enum ActiveFeature {
            case chat, voice, none
        }
    
        func switchTo(_ feature: ActiveFeature) async throws {
            switch feature {
            case .chat:
                // Only need LLM
                try? await RunAnywhere.unloadSTTModel()
                try? await RunAnywhere.unloadTTSVoice()
    
            case .voice:
                // Need all voice components
                try await RunAnywhere.loadSTTModel("whisper-base-onnx")
                try await RunAnywhere.loadTTSVoice("piper-en-us-amy")
    
            case .none:
                // Free all memory
                try? await RunAnywhere.unloadModel()
                try? await RunAnywhere.unloadSTTModel()
                try? await RunAnywhere.unloadTTSVoice()
            }
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/best-practices#monitor-memory)

Monitor Memory

Copy

Ask AI

    func checkMemoryBeforeOperation() -> Bool {
        let memoryInfo = ProcessInfo.processInfo
        let available = memoryInfo.physicalMemory
    
        // Require at least 2GB free for LLM operations
        return available > 2_000_000_000
    }
    

### 

[​](https://docs.runanywhere.ai/swift/best-practices#handle-memory-warnings)

Handle Memory Warnings

Copy

Ask AI

    class AppDelegate: NSObject, UIApplicationDelegate {
        func applicationDidReceiveMemoryWarning(_ application: UIApplication) {
            Task {
                // Unload non-essential models
                try? await RunAnywhere.unloadSTTModel()
                try? await RunAnywhere.unloadTTSVoice()
                try? await RunAnywhere.cleanTempFiles()
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/best-practices#model-selection)

Model Selection
----------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/best-practices#choose-appropriate-model-sizes)

Choose Appropriate Model Sizes

| Device | Recommended LLM | Notes |
| --- | --- | --- |
| iPhone 12/13 (4GB) | 1B Q4 | May need to unload others |
| iPhone 14/15 (6GB) | 1-3B Q4 | Good for most use cases |
| iPhone 15 Pro (8GB) | 3B Q4, some 7B | More headroom |
| iPad Pro | 3-7B Q4 | Depends on RAM |
| M1+ Mac | 7B+ Q4 | Ample memory |

### 

[​](https://docs.runanywhere.ai/swift/best-practices#device-specific-loading)

Device-Specific Loading

Copy

Ask AI

    func selectModelForDevice() -> String {
        let memory = ProcessInfo.processInfo.physicalMemory
    
        switch memory {
        case ..<4_000_000_000:
            return "llama-3.2-1b-instruct-q4"  // 1B for 4GB devices
        case 4_000_000_000..<8_000_000_000:
            return "llama-3.2-3b-instruct-q4"  // 3B for 4-8GB
        default:
            return "llama-3.2-7b-instruct-q4"  // 7B for 8GB+
        }
    }
    

[​](https://docs.runanywhere.ai/swift/best-practices#streaming-for-responsiveness)

Streaming for Responsiveness
------------------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/best-practices#always-stream-for-long-outputs)

Always Stream for Long Outputs

Copy

Ask AI

    // ❌ User waits for entire response
    let result = try await RunAnywhere.generate(prompt, options: LLMGenerationOptions(maxTokens: 500))
    
    // ✅ User sees tokens immediately
    let result = try await RunAnywhere.generateStream(prompt)
    for try await token in result.stream {
        updateUI(with: token)
    }
    

### 

[​](https://docs.runanywhere.ai/swift/best-practices#batch-ui-updates)

Batch UI Updates

Copy

Ask AI

    func streamWithBatching() async throws {
        let result = try await RunAnywhere.generateStream(prompt)
    
        var buffer = ""
        var lastUpdate = Date()
    
        for try await token in result.stream {
            buffer += token
    
            // Update UI every 50ms instead of every token
            if Date().timeIntervalSince(lastUpdate) > 0.05 {
                await MainActor.run {
                    self.displayText += buffer
                }
                buffer = ""
                lastUpdate = Date()
            }
        }
    
        // Flush remaining
        if !buffer.isEmpty {
            await MainActor.run {
                self.displayText += buffer
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/best-practices#threading)

Threading
----------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/best-practices#use-appropriate-actors)

Use Appropriate Actors

Copy

Ask AI

    // ✅ UI updates on MainActor
    await MainActor.run {
        self.response = result.text
    }
    
    // ✅ Heavy operations stay off main thread
    Task.detached {
        try await RunAnywhere.loadModel(modelId)
    }
    

### 

[​](https://docs.runanywhere.ai/swift/best-practices#don%E2%80%99t-block-the-main-thread)

Don’t Block the Main Thread

Copy

Ask AI

    // ❌ Blocking main thread
    @MainActor
    func loadModelSync() {
        // This blocks UI!
        RunLoop.current.run(until: Date(timeIntervalSinceNow: 5))
    }
    
    // ✅ Async loading with UI feedback
    @MainActor
    func loadModelAsync() {
        isLoading = true
    
        Task {
            try await RunAnywhere.loadModel(modelId)
            await MainActor.run {
                isLoading = false
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/best-practices#error-recovery)

Error Recovery
--------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/best-practices#implement-retries)

Implement Retries

Copy

Ask AI

    func generateWithRetry(prompt: String, maxAttempts: Int = 3) async throws -> String {
        for attempt in 1...maxAttempts {
            do {
                return try await RunAnywhere.chat(prompt)
            } catch let error as SDKError where error.code == .timeout {
                if attempt < maxAttempts {
                    try await Task.sleep(for: .seconds(1))
                    continue
                }
            }
        }
        throw SDKError.general(.timeout, "Max retries exceeded")
    }
    

### 

[​](https://docs.runanywhere.ai/swift/best-practices#graceful-fallbacks)

Graceful Fallbacks

Copy

Ask AI

    func speak(_ text: String) async {
        do {
            // Try neural voice
            try await RunAnywhere.speak(text)
        } catch {
            // Fall back to system voice
            let synthesizer = AVSpeechSynthesizer()
            let utterance = AVSpeechUtterance(string: text)
            synthesizer.speak(utterance)
        }
    }
    

[​](https://docs.runanywhere.ai/swift/best-practices#storage)

Storage
------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/best-practices#check-before-downloads)

Check Before Downloads

Copy

Ask AI

    func downloadModelSafely(_ model: ModelInfo) async throws {
        let storage = await RunAnywhere.getStorageInfo()
        let required = model.downloadSize ?? 0
    
        guard storage.availableBytes > required + 500_000_000 else {  // 500MB buffer
            throw SDKError.general(.insufficientStorage, "Not enough storage")
        }
    
        try await Download.shared.downloadModel(model)
    }
    

### 

[​](https://docs.runanywhere.ai/swift/best-practices#clean-up-regularly)

Clean Up Regularly

Copy

Ask AI

    func performMaintenance() async {
        // Clean temp files
        try? await RunAnywhere.cleanTempFiles()
    
        // Remove unused downloaded models
        let models = try? await RunAnywhere.availableModels()
        for model in models ?? [] {
            if model.isDownloaded && !isModelNeeded(model) {
                // Delete unused model
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/best-practices#event-handling)

Event Handling
--------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/best-practices#subscribe-to-events)

Subscribe to Events

Copy

Ask AI

    import Combine
    
    class AIViewModel: ObservableObject {
        @Published var isGenerating = false
        @Published var tokensPerSecond = 0.0
    
        private var cancellables = Set<AnyCancellable>()
    
        init() {
            RunAnywhere.events.events(for: .llm)
                .receive(on: DispatchQueue.main)
                .sink { [weak self] event in
                    self?.handleLLMEvent(event)
                }
                .store(in: &cancellables)
        }
    
        private func handleLLMEvent(_ event: SDKEvent) {
            switch event.type {
            case "generation.started":
                isGenerating = true
            case "generation.completed":
                isGenerating = false
            default:
                break
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/best-practices#testing)

Testing
------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/best-practices#use-development-environment)

Use Development Environment

Copy

Ask AI

    #if DEBUG
    try RunAnywhere.initialize(environment: .development)
    RunAnywhere.setLogLevel(.debug)
    #endif
    

### 

[​](https://docs.runanywhere.ai/swift/best-practices#mock-for-unit-tests)

Mock for Unit Tests

Copy

Ask AI

    protocol AIService {
        func generate(_ prompt: String) async throws -> String
    }
    
    class RealAIService: AIService {
        func generate(_ prompt: String) async throws -> String {
            try await RunAnywhere.chat(prompt)
        }
    }
    
    class MockAIService: AIService {
        func generate(_ prompt: String) async throws -> String {
            return "Mock response for: \(prompt)"
        }
    }
    

[​](https://docs.runanywhere.ai/swift/best-practices#quick-reference)

Quick Reference
----------------------------------------------------------------------------------------

| Scenario | Recommendation |
| --- | --- |
| App launch | Initialize SDK, register modules |
| Before first use | Pre-load models |
| Long text generation | Use streaming |
| Low memory devices | Use smaller models, unload when done |
| Voice features | Use Voice Agent for full pipeline |
| Production | Set log level to `.warning` |
| Errors | Implement retries and fallbacks |
| Background | Unload models to free memory |

[Configuration\
-------------\
\
SDK configuration →](https://docs.runanywhere.ai/swift/configuration)
[Error Handling\
--------------\
\
Handle errors gracefully →](https://docs.runanywhere.ai/swift/error-handling)

[Error Handling](https://docs.runanywhere.ai/swift/error-handling)
[Introduction](https://docs.runanywhere.ai/kotlin/introduction)

⌘I

---

# generate() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/llm/generate#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

generate()

On this page

*   [Basic Usage](https://docs.runanywhere.ai/swift/llm/generate#basic-usage)
    
*   [Method Signature](https://docs.runanywhere.ai/swift/llm/generate#method-signature)
    
*   [Parameters](https://docs.runanywhere.ai/swift/llm/generate#parameters)
    
*   [Returns](https://docs.runanywhere.ai/swift/llm/generate#returns)
    
*   [LLMGenerationResult](https://docs.runanywhere.ai/swift/llm/generate#llmgenerationresult)
    
*   [LLMGenerationOptions](https://docs.runanywhere.ai/swift/llm/generate#llmgenerationoptions)
    
*   [Generation Parameters](https://docs.runanywhere.ai/swift/llm/generate#generation-parameters)
    
*   [Examples](https://docs.runanywhere.ai/swift/llm/generate#examples)
    
*   [Basic Generation](https://docs.runanywhere.ai/swift/llm/generate#basic-generation)
    
*   [With Custom Options](https://docs.runanywhere.ai/swift/llm/generate#with-custom-options)
    
*   [With System Prompt](https://docs.runanywhere.ai/swift/llm/generate#with-system-prompt)
    
*   [For Reasoning Models](https://docs.runanywhere.ai/swift/llm/generate#for-reasoning-models)
    
*   [Performance Monitoring](https://docs.runanywhere.ai/swift/llm/generate#performance-monitoring)
    
*   [Structured Output](https://docs.runanywhere.ai/swift/llm/generate#structured-output)
    
*   [Error Handling](https://docs.runanywhere.ai/swift/llm/generate#error-handling)
    
*   [Temperature Guide](https://docs.runanywhere.ai/swift/llm/generate#temperature-guide)
    

The `generate()` method provides complete control over text generation with detailed performance metrics and customizable options.

[​](https://docs.runanywhere.ai/swift/llm/generate#basic-usage)

Basic Usage
------------------------------------------------------------------------------

Copy

Ask AI

    let result = try await RunAnywhere.generate(
        "Explain quantum computing in simple terms",
        options: LLMGenerationOptions(
            maxTokens: 200,
            temperature: 0.7
        )
    )
    
    print("Response: \(result.text)")
    print("Speed: \(result.tokensPerSecond) tok/s")
    

[​](https://docs.runanywhere.ai/swift/llm/generate#method-signature)

Method Signature
----------------------------------------------------------------------------------------

Copy

Ask AI

    public static func generate(
        _ prompt: String,
        options: LLMGenerationOptions? = nil
    ) async throws -> LLMGenerationResult
    

### 

[​](https://docs.runanywhere.ai/swift/llm/generate#parameters)

Parameters

| Parameter | Type | Description |
| --- | --- | --- |
| `prompt` | `String` | The text prompt |
| `options` | `LLMGenerationOptions?` | Generation configuration (optional) |

### 

[​](https://docs.runanywhere.ai/swift/llm/generate#returns)

Returns

An `LLMGenerationResult` containing the response and metrics.

[​](https://docs.runanywhere.ai/swift/llm/generate#llmgenerationresult)

LLMGenerationResult
----------------------------------------------------------------------------------------------

Copy

Ask AI

    public struct LLMGenerationResult: Sendable {
        public let text: String              // Generated text
        public let thinkingContent: String?  // Reasoning tokens (for thinking models)
        public let inputTokens: Int          // Prompt tokens
        public let tokensUsed: Int           // Total output tokens
        public let modelUsed: String         // Model ID
        public let latencyMs: TimeInterval   // Total generation time
        public let framework: String?        // Backend framework used
        public let tokensPerSecond: Double   // Generation speed
        public let timeToFirstTokenMs: Double? // Time to first token
        public let thinkingTokens: Int?      // Reasoning token count
        public let responseTokens: Int       // Response token count
    }
    

[​](https://docs.runanywhere.ai/swift/llm/generate#llmgenerationoptions)

LLMGenerationOptions
------------------------------------------------------------------------------------------------

Copy

Ask AI

    let options = LLMGenerationOptions(
        maxTokens: 100,           // Maximum tokens to generate (default: 100)
        temperature: 0.8,         // Randomness 0.0-2.0 (default: 0.8)
        topP: 1.0,                // Nucleus sampling (default: 1.0)
        stopSequences: ["###"],   // Stop generation at these strings
        streamingEnabled: false,  // Enable token streaming
        preferredFramework: .llamaCpp,  // Preferred backend
        systemPrompt: "You are a helpful assistant."
    )
    

### 

[​](https://docs.runanywhere.ai/swift/llm/generate#generation-parameters)

Generation Parameters

| Parameter | Type | Default | Description |
| --- | --- | --- | --- |
| `maxTokens` | `Int` | 100 | Maximum tokens to generate |
| `temperature` | `Float` | 0.8 | Controls randomness (0.0 = deterministic, 2.0 = very random) |
| `topP` | `Float` | 1.0 | Nucleus sampling threshold |
| `stopSequences` | `[String]` | `[]` | Stop generation at these strings |
| `streamingEnabled` | `Bool` | false | Enable token-by-token streaming |
| `preferredFramework` | `InferenceFramework?` | nil | Preferred backend framework |
| `systemPrompt` | `String?` | nil | System prompt for behavior |

[​](https://docs.runanywhere.ai/swift/llm/generate#examples)

Examples
------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/llm/generate#basic-generation)

Basic Generation

Copy

Ask AI

    let result = try await RunAnywhere.generate("Write a haiku about programming")
    print(result.text)
    print("Generated in \(result.latencyMs)ms")
    

### 

[​](https://docs.runanywhere.ai/swift/llm/generate#with-custom-options)

With Custom Options

Copy

Ask AI

    let result = try await RunAnywhere.generate(
        "Write a creative story about a robot",
        options: LLMGenerationOptions(
            maxTokens: 500,
            temperature: 1.2,  // More creative
            topP: 0.9,
            stopSequences: ["THE END"]
        )
    )
    

### 

[​](https://docs.runanywhere.ai/swift/llm/generate#with-system-prompt)

With System Prompt

Copy

Ask AI

    let result = try await RunAnywhere.generate(
        "What should I cook tonight?",
        options: LLMGenerationOptions(
            maxTokens: 200,
            systemPrompt: "You are a professional chef. Suggest creative recipes with detailed instructions."
        )
    )
    

### 

[​](https://docs.runanywhere.ai/swift/llm/generate#for-reasoning-models)

For Reasoning Models

Some models output their reasoning process. Extract it with `thinkingContent`:

Copy

Ask AI

    let result = try await RunAnywhere.generate(
        "Solve: If a train travels 60 mph for 2 hours, how far does it go?",
        options: LLMGenerationOptions(maxTokens: 300)
    )
    
    // The main response
    print("Answer: \(result.text)")
    
    // The model's reasoning (if available)
    if let thinking = result.thinkingContent {
        print("Reasoning: \(thinking)")
    }
    
    // Token breakdown
    if let thinkingTokens = result.thinkingTokens {
        print("Thinking tokens: \(thinkingTokens)")
    }
    print("Response tokens: \(result.responseTokens)")
    

### 

[​](https://docs.runanywhere.ai/swift/llm/generate#performance-monitoring)

Performance Monitoring

Copy

Ask AI

    let result = try await RunAnywhere.generate(prompt)
    
    // Performance metrics
    print("Model: \(result.modelUsed)")
    print("Input tokens: \(result.inputTokens)")
    print("Output tokens: \(result.tokensUsed)")
    print("Speed: \(String(format: "%.1f", result.tokensPerSecond)) tok/s")
    print("Latency: \(String(format: "%.0f", result.latencyMs))ms")
    
    if let ttft = result.timeToFirstTokenMs {
        print("Time to first token: \(String(format: "%.0f", ttft))ms")
    }
    

[​](https://docs.runanywhere.ai/swift/llm/generate#structured-output)

Structured Output
------------------------------------------------------------------------------------------

Generate type-safe structured output using the `Generatable` protocol:

Copy

Ask AI

    struct Recipe: Generatable {
        let name: String
        let ingredients: [String]
        let steps: [String]
        let cookingTime: Int
    
        static var jsonSchema: String {
            """
            {
              "type": "object",
              "properties": {
                "name": { "type": "string" },
                "ingredients": { "type": "array", "items": { "type": "string" } },
                "steps": { "type": "array", "items": { "type": "string" } },
                "cookingTime": { "type": "integer" }
              },
              "required": ["name", "ingredients", "steps", "cookingTime"]
            }
            """
        }
    }
    
    let recipe: Recipe = try await RunAnywhere.generateStructured(
        Recipe.self,
        prompt: "Create a simple pasta recipe"
    )
    
    print("Recipe: \(recipe.name)")
    print("Ingredients: \(recipe.ingredients.joined(separator: ", "))")
    print("Cook time: \(recipe.cookingTime) minutes")
    

[​](https://docs.runanywhere.ai/swift/llm/generate#error-handling)

Error Handling
------------------------------------------------------------------------------------

Copy

Ask AI

    do {
        let result = try await RunAnywhere.generate(prompt, options: options)
        print(result.text)
    } catch let error as SDKError {
        switch error.code {
        case .notInitialized:
            print("SDK not initialized")
        case .modelNotFound:
            print("Model not loaded")
        case .generationFailed:
            print("Generation failed: \(error.message)")
        case .contextTooLong:
            print("Prompt too long for model's context window")
        default:
            print("Error: \(error.localizedDescription)")
        }
    }
    

[​](https://docs.runanywhere.ai/swift/llm/generate#temperature-guide)

Temperature Guide
------------------------------------------------------------------------------------------

| Temperature | Use Case |
| --- | --- |
| 0.0 | Deterministic, factual answers |
| 0.3-0.5 | Focused, coherent responses |
| 0.7-0.8 | Balanced creativity (default) |
| 1.0-1.2 | Creative writing, brainstorming |
| 1.5+ | Very random, experimental |

[generateStream()\
----------------\
\
Stream tokens in real-time →](https://docs.runanywhere.ai/swift/llm/stream)
[System Prompts\
--------------\
\
Control model behavior →](https://docs.runanywhere.ai/swift/llm/system-prompts)

[chat()](https://docs.runanywhere.ai/swift/llm/chat)
[generateStream()](https://docs.runanywhere.ai/swift/llm/stream)

⌘I

---

# generate() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/llm/generate#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

generate()

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/llm/generate#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/react-native/llm/generate#basic-usage)
    
*   [API Reference](https://docs.runanywhere.ai/react-native/llm/generate#api-reference)
    
*   [Parameters](https://docs.runanywhere.ai/react-native/llm/generate#parameters)
    
*   [Returns](https://docs.runanywhere.ai/react-native/llm/generate#returns)
    
*   [Generation Options](https://docs.runanywhere.ai/react-native/llm/generate#generation-options)
    
*   [Temperature](https://docs.runanywhere.ai/react-native/llm/generate#temperature)
    
*   [Max Tokens](https://docs.runanywhere.ai/react-native/llm/generate#max-tokens)
    
*   [Stop Sequences](https://docs.runanywhere.ai/react-native/llm/generate#stop-sequences)
    
*   [System Prompts](https://docs.runanywhere.ai/react-native/llm/generate#system-prompts)
    
*   [Examples](https://docs.runanywhere.ai/react-native/llm/generate#examples)
    
*   [Full Example with Metrics](https://docs.runanywhere.ai/react-native/llm/generate#full-example-with-metrics)
    
*   [React Hook](https://docs.runanywhere.ai/react-native/llm/generate#react-hook)
    
*   [Thinking Models](https://docs.runanywhere.ai/react-native/llm/generate#thinking-models)
    
*   [Cancellation](https://docs.runanywhere.ai/react-native/llm/generate#cancellation)
    
*   [Related](https://docs.runanywhere.ai/react-native/llm/generate#related)
    

[​](https://docs.runanywhere.ai/react-native/llm/generate#overview)

Overview
-------------------------------------------------------------------------------

The `generate()` method provides full control over text generation with customizable options and detailed performance metrics. Use this for production applications where you need fine-grained control.

[​](https://docs.runanywhere.ai/react-native/llm/generate#basic-usage)

Basic Usage
-------------------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    const result = await RunAnywhere.generate('Explain quantum computing in simple terms', {
      maxTokens: 200,
      temperature: 0.7,
    })
    
    console.log('Response:', result.text)
    console.log('Tokens used:', result.tokensUsed)
    console.log('Speed:', result.performanceMetrics.tokensPerSecond, 'tok/s')
    console.log('Latency:', result.latencyMs, 'ms')
    

[​](https://docs.runanywhere.ai/react-native/llm/generate#api-reference)

API Reference
-----------------------------------------------------------------------------------------

Copy

Ask AI

    await RunAnywhere.generate(
      prompt: string,
      options?: GenerationOptions
    ): Promise<GenerationResult>
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/generate#parameters)

Parameters

Copy

Ask AI

    interface GenerationOptions {
      /** Maximum tokens to generate (default: 256) */
      maxTokens?: number
    
      /** Sampling temperature 0.0–2.0 (default: 0.7) */
      temperature?: number
    
      /** Top-p nucleus sampling (default: 0.95) */
      topP?: number
    
      /** Stop generation at these sequences */
      stopSequences?: string[]
    
      /** System prompt to define AI behavior */
      systemPrompt?: string
    
      /** Preferred execution target */
      preferredExecutionTarget?: ExecutionTarget
    
      /** Preferred inference framework */
      preferredFramework?: LLMFramework
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/generate#returns)

Returns

Copy

Ask AI

    interface GenerationResult {
      /** Generated text (thinking content removed if extracted) */
      text: string
    
      /** Extracted thinking/reasoning content (if model supports it) */
      thinkingContent?: string
    
      /** Total tokens used (prompt + response) */
      tokensUsed: number
    
      /** Number of tokens in the response */
      responseTokens: number
    
      /** Model ID that was used */
      modelUsed: string
    
      /** Total latency in milliseconds */
      latencyMs: number
    
      /** Execution target (onDevice/cloud/hybrid) */
      executionTarget: ExecutionTarget
    
      /** Framework used for inference */
      framework?: LLMFramework
    
      /** Hardware acceleration used */
      hardwareUsed: HardwareAcceleration
    
      /** Memory used during generation (bytes) */
      memoryUsed: number
    
      /** Detailed performance metrics */
      performanceMetrics: PerformanceMetrics
    }
    
    interface PerformanceMetrics {
      /** Time to first token in ms */
      timeToFirstTokenMs?: number
    
      /** Tokens generated per second */
      tokensPerSecond?: number
    
      /** Total inference time in ms */
      inferenceTimeMs: number
    }
    

[​](https://docs.runanywhere.ai/react-native/llm/generate#generation-options)

Generation Options
---------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/llm/generate#temperature)

Temperature

Controls randomness in the output. Lower values make output more focused and deterministic.

Copy

Ask AI

    // Creative writing - higher temperature
    const creative = await RunAnywhere.generate('Write a poem about the ocean', {
      temperature: 1.2,
      maxTokens: 150,
    })
    
    // Factual response - lower temperature
    const factual = await RunAnywhere.generate('What is the boiling point of water?', {
      temperature: 0.1,
      maxTokens: 50,
    })
    
    // Balanced (default)
    const balanced = await RunAnywhere.generate('Explain machine learning', {
      temperature: 0.7,
      maxTokens: 200,
    })
    

| Temperature | Use Case |
| --- | --- |
| 0.0–0.3 | Factual, deterministic responses |
| 0.4–0.7 | Balanced, general-purpose |
| 0.8–1.2 | Creative, varied outputs |
| 1.3–2.0 | Very creative, experimental |

### 

[​](https://docs.runanywhere.ai/react-native/llm/generate#max-tokens)

Max Tokens

Limits the length of the generated response.

Copy

Ask AI

    // Short answer
    const short = await RunAnywhere.generate('What is 2+2?', { maxTokens: 10 })
    
    // Detailed explanation
    const detailed = await RunAnywhere.generate('Explain how computers work', { maxTokens: 500 })
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/generate#stop-sequences)

Stop Sequences

Stop generation when specific sequences are encountered.

Copy

Ask AI

    const result = await RunAnywhere.generate('List 3 fruits:', {
      maxTokens: 100,
      stopSequences: ['4.', '\n\n'], // Stop at 4th item or double newline
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/generate#system-prompts)

System Prompts

Define the AI’s behavior and persona.

Copy

Ask AI

    const result = await RunAnywhere.generate('What is the best programming language?', {
      maxTokens: 200,
      systemPrompt: 'You are a helpful coding assistant. Be concise and practical.',
    })
    

See [System Prompts](https://docs.runanywhere.ai/react-native/llm/system-prompts)
 for more details.

[​](https://docs.runanywhere.ai/react-native/llm/generate#examples)

Examples
-------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/llm/generate#full-example-with-metrics)

Full Example with Metrics

Copy

Ask AI

    async function generateWithMetrics(prompt: string) {
      const result = await RunAnywhere.generate(prompt, {
        maxTokens: 200,
        temperature: 0.7,
      })
    
      console.log('=== Generation Results ===')
      console.log('Response:', result.text)
      console.log('')
      console.log('=== Metrics ===')
      console.log('Tokens used:', result.tokensUsed)
      console.log('Response tokens:', result.responseTokens)
      console.log('Total latency:', result.latencyMs, 'ms')
      console.log('TTFT:', result.performanceMetrics.timeToFirstTokenMs, 'ms')
      console.log('Speed:', result.performanceMetrics.tokensPerSecond?.toFixed(1), 'tok/s')
      console.log('Hardware:', result.hardwareUsed)
      console.log('Memory used:', (result.memoryUsed / 1024 / 1024).toFixed(1), 'MB')
    
      return result
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/generate#react-hook)

React Hook

useGenerate.ts

Copy

Ask AI

    import { useState, useCallback } from 'react'
    import { RunAnywhere, GenerationOptions, GenerationResult } from '@runanywhere/core'
    
    export function useGenerate() {
      const [result, setResult] = useState<GenerationResult | null>(null)
      const [isLoading, setIsLoading] = useState(false)
      const [error, setError] = useState<Error | null>(null)
    
      const generate = useCallback(async (prompt: string, options?: GenerationOptions) => {
        setIsLoading(true)
        setError(null)
    
        try {
          const res = await RunAnywhere.generate(prompt, options)
          setResult(res)
          return res
        } catch (err) {
          const e = err instanceof Error ? err : new Error('Generation failed')
          setError(e)
          throw e
        } finally {
          setIsLoading(false)
        }
      }, [])
    
      const reset = useCallback(() => {
        setResult(null)
        setError(null)
      }, [])
    
      return { generate, result, isLoading, error, reset }
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/generate#thinking-models)

Thinking Models

Some models support “thinking” or reasoning before responding:

Copy

Ask AI

    // Add a model with thinking support
    await LlamaCPP.addModel({
      id: 'qwq-32b',
      name: 'QwQ 32B',
      url: 'https://huggingface.co/.../qwq-32b-q4_k_m.gguf',
      memoryRequirement: 20_000_000_000,
      supportsThinking: true, // Enable thinking extraction
    })
    
    // Generate with thinking
    const result = await RunAnywhere.generate('Solve this step by step: What is 15% of 240?', {
      maxTokens: 500,
    })
    
    console.log('Thinking:', result.thinkingContent)
    // "Let me calculate 15% of 240. First, I'll convert 15% to a decimal..."
    
    console.log('Answer:', result.text)
    // "15% of 240 is 36."
    

[​](https://docs.runanywhere.ai/react-native/llm/generate#cancellation)

Cancellation
---------------------------------------------------------------------------------------

Cancel an ongoing generation:

Copy

Ask AI

    // Start generation
    const promise = RunAnywhere.generate('Write a long story...', { maxTokens: 1000 })
    
    // Cancel after 2 seconds
    setTimeout(() => {
      RunAnywhere.cancelGeneration()
    }, 2000)
    
    try {
      const result = await promise
    } catch (error) {
      if (isSDKError(error) && error.code === SDKErrorCode.generationCancelled) {
        console.log('Generation was cancelled')
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/llm/generate#related)

Related
-----------------------------------------------------------------------------

[Chat\
----\
\
Simple one-liner interface](https://docs.runanywhere.ai/react-native/llm/chat)
[Streaming\
---------\
\
Real-time token streaming](https://docs.runanywhere.ai/react-native/llm/stream)
[System Prompts\
--------------\
\
Control AI behavior](https://docs.runanywhere.ai/react-native/llm/system-prompts)
[Best Practices\
--------------\
\
Optimization tips](https://docs.runanywhere.ai/react-native/best-practices)

[chat()](https://docs.runanywhere.ai/react-native/llm/chat)
[generateStream()](https://docs.runanywhere.ai/react-native/llm/stream)

⌘I

---

# Quick Start - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/quick-start#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

React Native SDK

Quick Start

On this page

*   [Complete Example](https://docs.runanywhere.ai/react-native/quick-start#complete-example)
    
*   [Step-by-Step Guide](https://docs.runanywhere.ai/react-native/quick-start#step-by-step-guide)
    
*   [1\. Initialize the SDK](https://docs.runanywhere.ai/react-native/quick-start#1-initialize-the-sdk)
    
*   [Environment Options](https://docs.runanywhere.ai/react-native/quick-start#environment-options)
    
*   [2\. Register Backend Modules](https://docs.runanywhere.ai/react-native/quick-start#2-register-backend-modules)
    
*   [3\. Add Models](https://docs.runanywhere.ai/react-native/quick-start#3-add-models)
    
*   [4\. Download the Model](https://docs.runanywhere.ai/react-native/quick-start#4-download-the-model)
    
*   [5\. Load and Generate](https://docs.runanywhere.ai/react-native/quick-start#5-load-and-generate)
    
*   [6\. Stream Responses](https://docs.runanywhere.ai/react-native/quick-start#6-stream-responses)
    
*   [Using in a React Component](https://docs.runanywhere.ai/react-native/quick-start#using-in-a-react-component)
    
*   [What’s Next?](https://docs.runanywhere.ai/react-native/quick-start#what%E2%80%99s-next)
    

[​](https://docs.runanywhere.ai/react-native/quick-start#complete-example)

Complete Example
----------------------------------------------------------------------------------------------

Here’s a complete example to get you started with on-device text generation:

App.tsx

Copy

Ask AI

    import React, { useState, useEffect } from 'react';
    import { View, Text, Button, ActivityIndicator } from 'react-native';
    import { RunAnywhere, SDKEnvironment, ModelCategory } from '@runanywhere/core';
    import { LlamaCPP } from '@runanywhere/llamacpp';
    
    export default function App() {
      const [isReady, setIsReady] = useState(false);
      const [response, setResponse] = useState('');
      const [loading, setLoading] = useState(false);
    
      useEffect(() => {
        async function setup() {
          // 1. Initialize SDK (fast, ~1-5ms)
          await RunAnywhere.initialize({
            environment: SDKEnvironment.Development,
          });
    
          // 2. Register LlamaCPP module
          LlamaCPP.register();
    
          // 3. Add a model
          await LlamaCPP.addModel({
            id: 'smollm2-360m',
            name: 'SmolLM2 360M',
            url: 'https://huggingface.co/prithivMLmods/SmolLM2-360M-GGUF/resolve/main/SmolLM2-360M.Q8_0.gguf',
            memoryRequirement: 500_000_000,
          });
    
          // 4. Download model (shows progress in console)
          await RunAnywhere.downloadModel('smollm2-360m', (progress) => {
            console.log(`Download: ${(progress.progress * 100).toFixed(1)}%`);
          });
    
          // 5. Load model into memory
          const modelInfo = await RunAnywhere.getModelInfo('smollm2-360m');
          if (modelInfo?.localPath) {
            await RunAnywhere.loadModel(modelInfo.localPath);
          }
    
          setIsReady(true);
        }
    
        setup();
      }, []);
    
      const generateResponse = async () => {
        setLoading(true);
        try {
          const result = await RunAnywhere.generate(
            'Explain quantum computing in simple terms',
            {
              maxTokens: 200,
              temperature: 0.7,
            }
          );
          setResponse(result.text);
          console.log(`Generated in ${result.latencyMs}ms at ${result.performanceMetrics.tokensPerSecond} tok/s`);
        } finally {
          setLoading(false);
        }
      };
    
      if (!isReady) {
        return (
          <View style={{ flex: 1, justifyContent: 'center', alignItems: 'center' }}>
            <ActivityIndicator size="large" />
            <Text>Setting up AI...</Text>
          </View>
        );
      }
    
      return (
        <View style={{ flex: 1, padding: 20 }}>
          <Button title="Generate" onPress={generateResponse} disabled={loading} />
          {loading && <ActivityIndicator />}
          <Text style={{ marginTop: 20 }}>{response}</Text>
        </View>
      );
    }
    

[​](https://docs.runanywhere.ai/react-native/quick-start#step-by-step-guide)

Step-by-Step Guide
--------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/quick-start#1-initialize-the-sdk)

1\. Initialize the SDK

Initialize RunAnywhere once when your app starts:

Copy

Ask AI

    import { RunAnywhere, SDKEnvironment } from '@runanywhere/core'
    
    await RunAnywhere.initialize({
      environment: SDKEnvironment.Development, // No API key needed
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/quick-start#environment-options)

Environment Options

| Environment | Log Level | Description |
| --- | --- | --- |
| `Development` | Debug | Full logging, local testing |
| `Staging` | Info | Staging backend, moderate logging |
| `Production` | Warning | Production backend, minimal logging |

### 

[​](https://docs.runanywhere.ai/react-native/quick-start#2-register-backend-modules)

2\. Register Backend Modules

Register the backend modules you need:

Copy

Ask AI

    import { LlamaCPP } from '@runanywhere/llamacpp'
    import { ONNX } from '@runanywhere/onnx'
    
    // Register LLM backend
    LlamaCPP.register()
    
    // Register STT/TTS backend (if needed)
    ONNX.register()
    

### 

[​](https://docs.runanywhere.ai/react-native/quick-start#3-add-models)

3\. Add Models

Add models from HuggingFace or any direct URL:

Copy

Ask AI

    // Add LLM model
    await LlamaCPP.addModel({
      id: 'smollm2-360m',
      name: 'SmolLM2 360M',
      url: 'https://huggingface.co/.../SmolLM2-360M.Q8_0.gguf',
      memoryRequirement: 500_000_000,
    })
    
    // Add STT model
    await ONNX.addModel({
      id: 'whisper-tiny-en',
      name: 'Whisper Tiny English',
      url: 'https://github.com/.../sherpa-onnx-whisper-tiny.en.tar.gz',
      modality: ModelCategory.SpeechRecognition,
      artifactType: ModelArtifactType.TarGzArchive,
      memoryRequirement: 75_000_000,
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/quick-start#4-download-the-model)

4\. Download the Model

Download with progress tracking:

Copy

Ask AI

    await RunAnywhere.downloadModel('smollm2-360m', (progress) => {
      switch (progress.state) {
        case 'downloading':
          const percent = (progress.progress * 100).toFixed(1)
          console.log(`Downloading: ${percent}%`)
          break
        case 'extracting':
          console.log('Extracting archive...')
          break
        case 'completed':
          console.log('Download complete!')
          break
        case 'failed':
          console.error('Download failed')
          break
      }
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/quick-start#5-load-and-generate)

5\. Load and Generate

Copy

Ask AI

    // Load the model into memory (loadModel takes a file path, not a model ID)
    const modelInfo = await RunAnywhere.getModelInfo('smollm2-360m')
    await RunAnywhere.loadModel(modelInfo.localPath)
    
    // Simple chat
    const response = await RunAnywhere.chat('What is 2+2?')
    console.log(response) // "4"
    
    // Or with full metrics
    const result = await RunAnywhere.generate('Write a haiku about coding', { maxTokens: 50 })
    console.log('Response:', result.text)
    console.log('Tokens/sec:', result.performanceMetrics.tokensPerSecond)
    

### 

[​](https://docs.runanywhere.ai/react-native/quick-start#6-stream-responses)

6\. Stream Responses

For real-time token streaming:

Copy

Ask AI

    const streamResult = await RunAnywhere.generateStream('Tell me a story about AI', {
      maxTokens: 200,
    })
    
    // Accumulate and display tokens as they arrive
    let fullResponse = ''
    for await (const token of streamResult.stream) {
      fullResponse += token
      setResponse(fullResponse) // Update UI state
    }
    
    // Get final metrics (LLMGenerationResult has tokensPerSecond at top level)
    const finalResult = await streamResult.result
    console.log('Speed:', finalResult.tokensPerSecond, 'tok/s')
    

[​](https://docs.runanywhere.ai/react-native/quick-start#using-in-a-react-component)

Using in a React Component
------------------------------------------------------------------------------------------------------------------

Here’s a pattern for using the SDK in React components with hooks:

useAI.ts

Copy

Ask AI

    import { useState, useCallback } from 'react'
    import { RunAnywhere, GenerationOptions, GenerationResult } from '@runanywhere/core'
    
    export function useAI() {
      const [isGenerating, setIsGenerating] = useState(false)
      const [error, setError] = useState<string | null>(null)
    
      const generate = useCallback(
        async (prompt: string, options?: GenerationOptions): Promise<GenerationResult | null> => {
          setIsGenerating(true)
          setError(null)
    
          try {
            return await RunAnywhere.generate(prompt, options)
          } catch (err) {
            setError(err instanceof Error ? err.message : 'Generation failed')
            return null
          } finally {
            setIsGenerating(false)
          }
        },
        []
      )
    
      const chat = useCallback(async (prompt: string): Promise<string> => {
        setIsGenerating(true)
        setError(null)
    
        try {
          return await RunAnywhere.chat(prompt)
        } catch (err) {
          setError(err instanceof Error ? err.message : 'Chat failed')
          return ''
        } finally {
          setIsGenerating(false)
        }
      }, [])
    
      return { generate, chat, isGenerating, error }
    }
    

[​](https://docs.runanywhere.ai/react-native/quick-start#what%E2%80%99s-next)

What’s Next?
---------------------------------------------------------------------------------------------

[LLM Chat\
--------\
\
Simple one-liner for quick responses](https://docs.runanywhere.ai/react-native/llm/chat)
[Streaming\
---------\
\
Real-time token streaming](https://docs.runanywhere.ai/react-native/llm/stream)
[Speech-to-Text\
--------------\
\
Transcribe audio to text](https://docs.runanywhere.ai/react-native/stt/transcribe)
[Text-to-Speech\
--------------\
\
Convert text to spoken audio](https://docs.runanywhere.ai/react-native/tts/synthesize)

[Installation](https://docs.runanywhere.ai/react-native/installation)
[chat()](https://docs.runanywhere.ai/react-native/llm/chat)

⌘I

---

# System Prompts - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/llm/system-prompts#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

System Prompts

On this page

*   [Basic Usage](https://docs.runanywhere.ai/swift/llm/system-prompts#basic-usage)
    
*   [How System Prompts Work](https://docs.runanywhere.ai/swift/llm/system-prompts#how-system-prompts-work)
    
*   [Examples](https://docs.runanywhere.ai/swift/llm/system-prompts#examples)
    
*   [Customer Support Bot](https://docs.runanywhere.ai/swift/llm/system-prompts#customer-support-bot)
    
*   [Code Assistant](https://docs.runanywhere.ai/swift/llm/system-prompts#code-assistant)
    
*   [Creative Writer](https://docs.runanywhere.ai/swift/llm/system-prompts#creative-writer)
    
*   [JSON Output](https://docs.runanywhere.ai/swift/llm/system-prompts#json-output)
    
*   [Language Tutor](https://docs.runanywhere.ai/swift/llm/system-prompts#language-tutor)
    
*   [Best Practices](https://docs.runanywhere.ai/swift/llm/system-prompts#best-practices)
    
*   [Be Specific](https://docs.runanywhere.ai/swift/llm/system-prompts#be-specific)
    
*   [Set Boundaries](https://docs.runanywhere.ai/swift/llm/system-prompts#set-boundaries)
    
*   [Control Output Format](https://docs.runanywhere.ai/swift/llm/system-prompts#control-output-format)
    
*   [Maintain Persona](https://docs.runanywhere.ai/swift/llm/system-prompts#maintain-persona)
    
*   [Reusable System Prompts](https://docs.runanywhere.ai/swift/llm/system-prompts#reusable-system-prompts)
    
*   [Tips](https://docs.runanywhere.ai/swift/llm/system-prompts#tips)
    

System prompts define the AI’s personality, behavior, and constraints. They’re processed before the user’s prompt and help ensure consistent responses.

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#basic-usage)

Basic Usage
------------------------------------------------------------------------------------

Copy

Ask AI

    let result = try await RunAnywhere.generate(
        "What's the weather like?",
        options: LLMGenerationOptions(
            systemPrompt: "You are a helpful weather assistant. Always include temperature in both Celsius and Fahrenheit."
        )
    )
    

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#how-system-prompts-work)

How System Prompts Work
------------------------------------------------------------------------------------------------------------

The system prompt is prepended to your conversation, establishing context for the model:

Copy

Ask AI

    [System]: You are a helpful assistant...
    [User]: What's the weather like?
    [Assistant]: (generated response)
    

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#examples)

Examples
------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#customer-support-bot)

Customer Support Bot

Copy

Ask AI

    let systemPrompt = """
    You are a friendly customer support agent for TechCorp.
    
    Guidelines:
    - Be polite and professional
    - Ask clarifying questions when needed
    - Offer solutions, not just sympathy
    - If you don't know something, say so
    - Never share customer data
    
    Products: SmartWidget Pro, CloudSync Basic, CloudSync Pro
    Support hours: 9 AM - 5 PM EST
    """
    
    let result = try await RunAnywhere.generate(
        "My SmartWidget won't connect to WiFi",
        options: LLMGenerationOptions(
            maxTokens: 300,
            systemPrompt: systemPrompt
        )
    )
    

### 

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#code-assistant)

Code Assistant

Copy

Ask AI

    let systemPrompt = """
    You are an expert Swift developer assistant.
    
    When writing code:
    - Use Swift 5.9+ syntax
    - Follow Apple's API Design Guidelines
    - Include comments for complex logic
    - Handle errors appropriately
    - Use async/await for asynchronous code
    
    When explaining code:
    - Be concise but thorough
    - Mention relevant Apple frameworks
    - Suggest best practices
    """
    
    let result = try await RunAnywhere.generate(
        "How do I make a network request in Swift?",
        options: LLMGenerationOptions(
            maxTokens: 500,
            systemPrompt: systemPrompt
        )
    )
    

### 

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#creative-writer)

Creative Writer

Copy

Ask AI

    let systemPrompt = """
    You are a creative writing assistant specializing in science fiction.
    
    Style guidelines:
    - Use vivid, sensory descriptions
    - Create compelling characters
    - Build tension and suspense
    - Avoid clichés
    - Write in third person unless asked otherwise
    
    Tone: Thoughtful, imaginative, slightly melancholic
    """
    
    let result = try await RunAnywhere.generate(
        "Write an opening paragraph about a colony ship",
        options: LLMGenerationOptions(
            maxTokens: 300,
            temperature: 1.0,  // Higher creativity
            systemPrompt: systemPrompt
        )
    )
    

### 

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#json-output)

JSON Output

Copy

Ask AI

    let systemPrompt = """
    You are a data extraction assistant.
    Always respond with valid JSON only.
    Do not include any text before or after the JSON.
    Do not use markdown code blocks.
    """
    
    let result = try await RunAnywhere.generate(
        "Extract: 'John Smith, age 30, lives in New York'",
        options: LLMGenerationOptions(
            maxTokens: 100,
            temperature: 0.1,  // Low temperature for consistency
            systemPrompt: systemPrompt
        )
    )
    // Expected: {"name": "John Smith", "age": 30, "city": "New York"}
    

### 

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#language-tutor)

Language Tutor

Copy

Ask AI

    let systemPrompt = """
    You are a Spanish language tutor for beginners.
    
    Instructions:
    - Respond in both Spanish and English
    - Correct mistakes gently
    - Explain grammar rules when relevant
    - Use simple vocabulary
    - Include pronunciation tips in brackets
    
    Format responses as:
    Spanish: [response in Spanish]
    English: [English translation]
    Note: [any grammar or pronunciation tips]
    """
    
    let result = try await RunAnywhere.generate(
        "How do I say 'I want to learn Spanish'?",
        options: LLMGenerationOptions(
            maxTokens: 200,
            systemPrompt: systemPrompt
        )
    )
    

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#best-practices)

Best Practices
------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#be-specific)

Be Specific

Copy

Ask AI

    // ❌ Too vague
    let vague = "Be helpful"
    
    // ✅ Specific and actionable
    let specific = """
    You are a cooking assistant.
    - Suggest recipes based on available ingredients
    - Include cooking times and temperatures
    - Mention dietary restrictions when asked
    - Provide substitutions for common allergens
    """
    

### 

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#set-boundaries)

Set Boundaries

Copy

Ask AI

    let systemPrompt = """
    You are a medical information assistant.
    
    IMPORTANT:
    - You provide general health information only
    - You are NOT a replacement for professional medical advice
    - Always recommend consulting a healthcare provider
    - Never diagnose conditions or prescribe treatments
    - If asked about emergencies, direct to 911
    """
    

### 

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#control-output-format)

Control Output Format

Copy

Ask AI

    let systemPrompt = """
    Format all responses as bullet points.
    Keep each point under 20 words.
    Use simple language (8th grade reading level).
    """
    

### 

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#maintain-persona)

Maintain Persona

Copy

Ask AI

    let systemPrompt = """
    You are Max, a friendly AI assistant with these traits:
    - Enthusiastic and positive
    - Uses occasional emojis (sparingly)
    - Admits when unsure
    - Makes relevant pop culture references
    - Signs off with "Happy to help! 🚀"
    """
    

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#reusable-system-prompts)

Reusable System Prompts
------------------------------------------------------------------------------------------------------------

Create a library of system prompts for different use cases:

Copy

Ask AI

    enum SystemPrompts {
        static let codeAssistant = """
        You are an expert programmer...
        """
    
        static let creativeWriter = """
        You are a creative writing assistant...
        """
    
        static let dataExtractor = """
        Extract data as JSON...
        """
    
        static func custom(role: String, guidelines: [String]) -> String {
            """
            You are \(role).
    
            Guidelines:
            \(guidelines.map { "- \($0)" }.joined(separator: "\n"))
            """
        }
    }
    
    // Usage
    let result = try await RunAnywhere.generate(
        prompt,
        options: LLMGenerationOptions(
            systemPrompt: SystemPrompts.codeAssistant
        )
    )
    

[​](https://docs.runanywhere.ai/swift/llm/system-prompts#tips)

Tips
----------------------------------------------------------------------

Keep it concise

Long system prompts consume context window tokens. Be specific but brief.

Test variations

Small changes in wording can significantly affect output. Test different phrasings.

Lower temperature for following instructions

Use `temperature: 0.3-0.5` when you need the model to strictly follow format instructions.

Combine with stop sequences

Use `stopSequences` to prevent the model from generating beyond your desired format.

[generate()\
----------\
\
Learn more about generation options →](https://docs.runanywhere.ai/swift/llm/generate)

[generateStream()](https://docs.runanywhere.ai/swift/llm/stream)
[transcribe()](https://docs.runanywhere.ai/swift/stt/transcribe)

⌘I

---

# Voices - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/tts/voices#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

Voices

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/tts/voices#overview)
    
*   [Voice Types](https://docs.runanywhere.ai/react-native/tts/voices#voice-types)
    
*   [Getting Available Voices](https://docs.runanywhere.ai/react-native/tts/voices#getting-available-voices)
    
*   [Voice Info Structure](https://docs.runanywhere.ai/react-native/tts/voices#voice-info-structure)
    
*   [Using Neural Voices (Piper)](https://docs.runanywhere.ai/react-native/tts/voices#using-neural-voices-piper)
    
*   [Add a Piper Voice](https://docs.runanywhere.ai/react-native/tts/voices#add-a-piper-voice)
    
*   [Available Piper Voices](https://docs.runanywhere.ai/react-native/tts/voices#available-piper-voices)
    
*   [Using System Voices](https://docs.runanywhere.ai/react-native/tts/voices#using-system-voices)
    
*   [iOS (AVSpeechSynthesizer)](https://docs.runanywhere.ai/react-native/tts/voices#ios-avspeechsynthesizer)
    
*   [Android (TextToSpeech)](https://docs.runanywhere.ai/react-native/tts/voices#android-texttospeech)
    
*   [Voice Selection Component](https://docs.runanywhere.ai/react-native/tts/voices#voice-selection-component)
    
*   [Voice Comparison](https://docs.runanywhere.ai/react-native/tts/voices#voice-comparison)
    
*   [Neural vs System Voices](https://docs.runanywhere.ai/react-native/tts/voices#neural-vs-system-voices)
    
*   [Loading Multiple Voices](https://docs.runanywhere.ai/react-native/tts/voices#loading-multiple-voices)
    
*   [Best Practices](https://docs.runanywhere.ai/react-native/tts/voices#best-practices)
    
*   [Voice Selection Strategy](https://docs.runanywhere.ai/react-native/tts/voices#voice-selection-strategy)
    
*   [Related](https://docs.runanywhere.ai/react-native/tts/voices#related)
    

[​](https://docs.runanywhere.ai/react-native/tts/voices#overview)

Overview
-----------------------------------------------------------------------------

The RunAnywhere SDK supports both neural TTS voices (Piper) and system voices (platform-native TTS). This guide covers available voices and how to configure them.

[​](https://docs.runanywhere.ai/react-native/tts/voices#voice-types)

Voice Types
-----------------------------------------------------------------------------------

| Type | Provider | Quality | Offline | Languages |
| --- | --- | --- | --- | --- |
| Neural | Piper TTS | High |     | Multi |
| System | iOS AVSpeech | Good |     | Multi |
| System | Android TTS | Good |     | Multi |

[​](https://docs.runanywhere.ai/react-native/tts/voices#getting-available-voices)

Getting Available Voices
-------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    // Get all available voices
    const voices = await RunAnywhere.availableTTSVoices()
    
    for (const voice of voices) {
      console.log(`${voice.id}`)
      console.log(`  Name: ${voice.name}`)
      console.log(`  Language: ${voice.language}`)
      console.log(`  Type: ${voice.type}`)
      console.log(`  Quality: ${voice.quality}`)
    }
    

[​](https://docs.runanywhere.ai/react-native/tts/voices#voice-info-structure)

Voice Info Structure
-----------------------------------------------------------------------------------------------------

Copy

Ask AI

    interface TTSVoiceInfo {
      /** Unique voice identifier */
      id: string
    
      /** Display name */
      name: string
    
      /** Language code (e.g., 'en-US') */
      language: string
    
      /** Voice type (neural, system) */
      type: 'neural' | 'system'
    
      /** Quality level */
      quality: 'low' | 'medium' | 'high'
    
      /** Gender (if available) */
      gender?: 'male' | 'female' | 'neutral'
    
      /** Whether voice requires download */
      requiresDownload: boolean
    
      /** Whether voice is ready to use */
      isAvailable: boolean
    }
    

[​](https://docs.runanywhere.ai/react-native/tts/voices#using-neural-voices-piper)

Using Neural Voices (Piper)
-----------------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/tts/voices#add-a-piper-voice)

Add a Piper Voice

Copy

Ask AI

    import { ONNX, ModelArtifactType } from '@runanywhere/onnx'
    import { ModelCategory } from '@runanywhere/core'
    
    // English (US) - Lessac voice
    await ONNX.addModel({
      id: 'piper-en-lessac',
      name: 'English (Lessac)',
      url: 'https://github.com/RunanywhereAI/sherpa-onnx/releases/.../vits-piper-en_US-lessac-medium.tar.gz',
      modality: ModelCategory.SpeechSynthesis,
      artifactType: ModelArtifactType.TarGzArchive,
      memoryRequirement: 65_000_000,
    })
    
    // Download and load
    await RunAnywhere.downloadModel('piper-en-lessac')
    const modelInfo = await RunAnywhere.getModelInfo('piper-en-lessac')
    await RunAnywhere.loadTTSModel(modelInfo.localPath, 'piper')
    
    // Use the voice
    const result = await RunAnywhere.synthesize('Hello, world!')
    

### 

[​](https://docs.runanywhere.ai/react-native/tts/voices#available-piper-voices)

Available Piper Voices

| Voice ID | Language | Quality | Size |
| --- | --- | --- | --- |
| `en_US-lessac-medium` | English (US) | High | ~65MB |
| `en_US-amy-medium` | English (US) | High | ~65MB |
| `en_GB-alba-medium` | English (UK) | High | ~65MB |
| `de_DE-thorsten-medium` | German | High | ~65MB |
| `fr_FR-upmc-medium` | French | High | ~65MB |
| `es_ES-sharvard-medium` | Spanish | High | ~65MB |

Piper voices provide the highest quality but require downloading model files. See [Piper TTS](https://github.com/rhasspy/piper)
 for more voices.

[​](https://docs.runanywhere.ai/react-native/tts/voices#using-system-voices)

Using System Voices
---------------------------------------------------------------------------------------------------

System voices use the platform’s built-in TTS engine and don’t require downloads.

### 

[​](https://docs.runanywhere.ai/react-native/tts/voices#ios-avspeechsynthesizer)

iOS (AVSpeechSynthesizer)

Copy

Ask AI

    // Use system TTS (no model loading needed)
    await RunAnywhere.speak('Hello from iOS!', {
      voice: 'com.apple.ttsbundle.Samantha-compact',
      rate: 1.0,
      pitch: 1.0,
    })
    
    // List iOS voices
    const iosVoices = (await RunAnywhere.availableTTSVoices()).filter((v) => v.type === 'system')
    

### 

[​](https://docs.runanywhere.ai/react-native/tts/voices#android-texttospeech)

Android (TextToSpeech)

Copy

Ask AI

    // Use system TTS
    await RunAnywhere.speak('Hello from Android!', {
      voice: 'en-us-x-tpf-local',
      rate: 1.0,
      pitch: 1.0,
    })
    
    // List Android voices
    const androidVoices = (await RunAnywhere.availableTTSVoices()).filter((v) => v.type === 'system')
    

[​](https://docs.runanywhere.ai/react-native/tts/voices#voice-selection-component)

Voice Selection Component
---------------------------------------------------------------------------------------------------------------

VoiceSelector.tsx

Copy

Ask AI

    import React, { useState, useEffect } from 'react'
    import { View, Text, FlatList, TouchableOpacity, StyleSheet } from 'react-native'
    import { RunAnywhere, TTSVoiceInfo } from '@runanywhere/core'
    
    export function VoiceSelector({ onSelect }: { onSelect: (voice: TTSVoiceInfo) => void }) {
      const [voices, setVoices] = useState<TTSVoiceInfo[]>([])
      const [selected, setSelected] = useState<string | null>(null)
    
      useEffect(() => {
        RunAnywhere.availableTTSVoices().then(setVoices)
      }, [])
    
      const handleSelect = (voice: TTSVoiceInfo) => {
        setSelected(voice.id)
        onSelect(voice)
      }
    
      const renderVoice = ({ item }: { item: TTSVoiceInfo }) => (
        <TouchableOpacity
          style={[styles.voiceItem, selected === item.id && styles.selected]}
          onPress={() => handleSelect(item)}
          disabled={!item.isAvailable}
        >
          <Text style={styles.voiceName}>{item.name}</Text>
          <Text style={styles.voiceInfo}>
            {item.language} • {item.type} • {item.quality}
          </Text>
          {!item.isAvailable && <Text style={styles.downloadRequired}>Download required</Text>}
        </TouchableOpacity>
      )
    
      return (
        <FlatList
          data={voices}
          renderItem={renderVoice}
          keyExtractor={(item) => item.id}
          style={styles.list}
        />
      )
    }
    
    const styles = StyleSheet.create({
      list: { flex: 1 },
      voiceItem: {
        padding: 16,
        borderBottomWidth: 1,
        borderBottomColor: '#eee',
      },
      selected: {
        backgroundColor: '#e3f2fd',
      },
      voiceName: {
        fontSize: 16,
        fontWeight: '600',
      },
      voiceInfo: {
        fontSize: 12,
        color: '#666',
        marginTop: 4,
      },
      downloadRequired: {
        fontSize: 12,
        color: '#f44336',
        marginTop: 4,
      },
    })
    

[​](https://docs.runanywhere.ai/react-native/tts/voices#voice-comparison)

Voice Comparison
---------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/tts/voices#neural-vs-system-voices)

Neural vs System Voices

Copy

Ask AI

    // Compare voice quality
    async function compareVoices(text: string) {
      // Neural voice (higher quality, requires download)
      const neural = await RunAnywhere.synthesize(text, {
        voice: 'piper-en-lessac',
      })
    
      // System voice (built-in, no download)
      await RunAnywhere.speak(text, {
        voice: 'com.apple.ttsbundle.Samantha-compact',
      })
    
      console.log('Neural duration:', neural.duration, 's')
    }
    

| Aspect | Neural (Piper) | System |
| --- | --- | --- |
| Quality | Very High | Good |
| Naturalness | Very Natural | Natural |
| Download Required | Yes | No  |
| Offline | Yes | Yes |
| Customization | Limited | Platform-dependent |
| Languages | Many | Many |

[​](https://docs.runanywhere.ai/react-native/tts/voices#loading-multiple-voices)

Loading Multiple Voices
-----------------------------------------------------------------------------------------------------------

Copy

Ask AI

    // Load voice for use
    await RunAnywhere.loadTTSVoice('piper-en-lessac')
    
    // Switch voices without reloading model
    const result1 = await RunAnywhere.synthesize('Hello', { voice: 'voice1' })
    const result2 = await RunAnywhere.synthesize('Hello', { voice: 'voice2' })
    

[​](https://docs.runanywhere.ai/react-native/tts/voices#best-practices)

Best Practices
-----------------------------------------------------------------------------------------

For production apps, pre-download neural voices during onboarding so they’re ready when needed.

### 

[​](https://docs.runanywhere.ai/react-native/tts/voices#voice-selection-strategy)

Voice Selection Strategy

Copy

Ask AI

    async function getBestVoice(language: string): Promise<TTSVoiceInfo | null> {
      const voices = await RunAnywhere.availableTTSVoices()
    
      // Prefer neural voices if available
      const neuralVoice = voices.find(
        (v) => v.language.startsWith(language) && v.type === 'neural' && v.isAvailable
      )
    
      if (neuralVoice) return neuralVoice
    
      // Fall back to system voice
      const systemVoice = voices.find(
        (v) => v.language.startsWith(language) && v.type === 'system' && v.isAvailable
      )
    
      return systemVoice || null
    }
    

[​](https://docs.runanywhere.ai/react-native/tts/voices#related)

Related
---------------------------------------------------------------------------

[Synthesize\
----------\
\
Basic synthesis](https://docs.runanywhere.ai/react-native/tts/synthesize)
[TTS Streaming\
-------------\
\
Streaming synthesis](https://docs.runanywhere.ai/react-native/tts/stream)
[Voice Agent\
-----------\
\
Full voice pipeline](https://docs.runanywhere.ai/react-native/voice-agent)

[Streaming TTS](https://docs.runanywhere.ai/react-native/tts/stream)
[Voice Activity Detection](https://docs.runanywhere.ai/react-native/vad)

⌘I

---

# transcribe() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/stt/transcribe#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

transcribe()

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/stt/transcribe#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/react-native/stt/transcribe#basic-usage)
    
*   [Setup](https://docs.runanywhere.ai/react-native/stt/transcribe#setup)
    
*   [API Reference](https://docs.runanywhere.ai/react-native/stt/transcribe#api-reference)
    
*   [transcribeFile](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribefile)
    
*   [transcribe](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribe)
    
*   [transcribeBuffer](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribebuffer)
    
*   [STT Options](https://docs.runanywhere.ai/react-native/stt/transcribe#stt-options)
    
*   [STT Result](https://docs.runanywhere.ai/react-native/stt/transcribe#stt-result)
    
*   [Examples](https://docs.runanywhere.ai/react-native/stt/transcribe#examples)
    
*   [Transcribe Base64 Audio](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribe-base64-audio)
    
*   [Transcribe Float32 Buffer](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribe-float32-buffer)
    
*   [React Native Recording + Transcription](https://docs.runanywhere.ai/react-native/stt/transcribe#react-native-recording-%2B-transcription)
    
*   [Transcribe Buffer from Microphone](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribe-buffer-from-microphone)
    
*   [Supported Audio Formats](https://docs.runanywhere.ai/react-native/stt/transcribe#supported-audio-formats)
    
*   [Available Models](https://docs.runanywhere.ai/react-native/stt/transcribe#available-models)
    
*   [Error Handling](https://docs.runanywhere.ai/react-native/stt/transcribe#error-handling)
    
*   [Related](https://docs.runanywhere.ai/react-native/stt/transcribe#related)
    

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#overview)

Overview
---------------------------------------------------------------------------------

The Speech-to-Text (STT) API allows you to transcribe audio files or raw audio data to text using on-device Whisper models. All transcription happens locally on the device for privacy and offline capability.

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#basic-usage)

Basic Usage
---------------------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    // Transcribe raw audio samples (most common approach)
    const result = await RunAnywhere.transcribeBuffer(audioSamples, {
      language: 'en',
    })
    
    console.log('Transcription:', result.text)
    

Or transcribe base64-encoded audio:

Copy

Ask AI

    // Transcribe base64-encoded float32 PCM audio
    const result = await RunAnywhere.transcribe(audioBase64, {
      language: 'en',
    })
    
    console.log('Transcription:', result.text)
    

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#setup)

Setup
---------------------------------------------------------------------------

Before transcribing, you need to download and load an STT model:

Copy

Ask AI

    import { RunAnywhere, ModelCategory } from '@runanywhere/core'
    import { ONNX, ModelArtifactType } from '@runanywhere/onnx'
    
    // 1. Initialize SDK and register ONNX backend
    await RunAnywhere.initialize({ environment: SDKEnvironment.Development })
    ONNX.register()
    
    // 2. Add Whisper model
    await ONNX.addModel({
      id: 'whisper-tiny-en',
      name: 'Whisper Tiny English',
      url: 'https://github.com/RunanywhereAI/sherpa-onnx/releases/download/runanywhere-models-v1/sherpa-onnx-whisper-tiny.en.tar.gz',
      modality: ModelCategory.SpeechRecognition,
      artifactType: ModelArtifactType.TarGzArchive,
      memoryRequirement: 75_000_000,
    })
    
    // 3. Download model
    await RunAnywhere.downloadModel('whisper-tiny-en', (progress) => {
      console.log(`Download: ${(progress.progress * 100).toFixed(1)}%`)
    })
    
    // 4. Load model
    const modelInfo = await RunAnywhere.getModelInfo('whisper-tiny-en')
    await RunAnywhere.loadSTTModel(modelInfo.localPath, 'whisper')
    

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#api-reference)

API Reference
-------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribefile)

`transcribeFile`

`transcribeFile()` is declared in the SDK but **not yet implemented**. Calling it will throw: `"transcribeFile not yet implemented with rac_* API"`. Use `transcribe()` or `transcribeBuffer()` instead.

Copy

Ask AI

    await RunAnywhere.transcribeFile(
      audioPath: string,
      options?: STTOptions
    ): Promise<STTResult>
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribe)

`transcribe`

Transcribe base64-encoded audio data.

Copy

Ask AI

    await RunAnywhere.transcribe(
      audioData: string,
      options?: STTOptions
    ): Promise<STTResult>
    

**Parameters:**

| Parameter | Type | Description |
| --- | --- | --- |
| `audioData` | `string` | Base64-encoded float32 PCM audio data |
| `options` | `STTOptions` | Optional transcription settings |

### 

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribebuffer)

`transcribeBuffer`

Transcribe raw float32 audio samples directly.

Copy

Ask AI

    await RunAnywhere.transcribeBuffer(
      samples: Float32Array,
      options?: STTOptions
    ): Promise<STTResult>
    

**Parameters:**

| Parameter | Type | Description |
| --- | --- | --- |
| `samples` | `Float32Array` | Float32 PCM audio samples |
| `options` | `STTOptions` | Optional transcription settings |

`transcribeBuffer` converts the Float32Array to base64 internally and calls the native `transcribe()` method. The sample rate defaults to 16000 Hz.

### 

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#stt-options)

STT Options

Copy

Ask AI

    interface STTOptions {
      /** Language code (e.g., 'en', 'es', 'fr') */
      language?: string
    
      /** Enable punctuation in output */
      punctuation?: boolean
    
      /** Enable speaker diarization */
      diarization?: boolean
    
      /** Enable word-level timestamps */
      wordTimestamps?: boolean
    
      /** Audio sample rate (default: 16000) */
      sampleRate?: number
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#stt-result)

STT Result

Copy

Ask AI

    interface STTResult {
      /** Main transcription text */
      text: string
    
      /** Segments with timing information */
      segments: STTSegment[]
    
      /** Detected language code */
      language?: string
    
      /** Overall confidence (0.0 - 1.0) */
      confidence: number
    
      /** Audio duration in seconds */
      duration: number
    
      /** Alternative transcriptions */
      alternatives: STTAlternative[]
    }
    
    interface STTSegment {
      /** Segment text */
      text: string
    
      /** Start time in seconds */
      startTime: number
    
      /** End time in seconds */
      endTime: number
    
      /** Speaker ID (if diarization enabled) */
      speakerId?: string
    
      /** Segment confidence */
      confidence: number
    }
    

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#examples)

Examples
---------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribe-base64-audio)

Transcribe Base64 Audio

Copy

Ask AI

    // Simple transcription from base64 float32 PCM
    const result = await RunAnywhere.transcribe(audioBase64, {
      language: 'en',
    })
    console.log(result.text)
    
    // With language hint
    const spanish = await RunAnywhere.transcribe(spanishAudioBase64, {
      language: 'es',
    })
    console.log(spanish.text)
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribe-float32-buffer)

Transcribe Float32 Buffer

Copy

Ask AI

    // Transcribe from Float32Array samples
    const samples = new Float32Array(audioBuffer)
    const result = await RunAnywhere.transcribeBuffer(samples, {
      language: 'en',
    })
    console.log(result.text)
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#react-native-recording-+-transcription)

React Native Recording + Transcription

Using the SDK’s built-in `RunAnywhere.Audio` API for recording:

VoiceRecorder.tsx

Copy

Ask AI

    import React, { useState, useCallback } from 'react'
    import { View, Button, Text } from 'react-native'
    import { RunAnywhere } from '@runanywhere/core'
    import RNFS from 'react-native-fs'
    
    export function VoiceRecorder() {
      const [isRecording, setIsRecording] = useState(false)
      const [transcription, setTranscription] = useState('')
    
      const startRecording = useCallback(async () => {
        setIsRecording(true)
        setTranscription('')
        await RunAnywhere.Audio.requestPermission()
        await RunAnywhere.Audio.startRecording()
      }, [])
    
      const stopRecording = useCallback(async () => {
        setIsRecording(false)
        const recording = await RunAnywhere.Audio.stopRecording()
        // recording.uri  - path to the recorded WAV file
        // recording.durationMs - duration in milliseconds
    
        // Read the file and transcribe
        try {
          const audioBase64 = await RNFS.readFile(recording.uri, 'base64')
          const result = await RunAnywhere.transcribe(audioBase64, {
            language: 'en',
          })
          setTranscription(result.text)
        } catch (error) {
          setTranscription('Error: ' + (error as Error).message)
        }
      }, [])
    
      return (
        <View style={{ padding: 16 }}>
          <Button
            title={isRecording ? 'Stop Recording' : 'Start Recording'}
            onPress={isRecording ? stopRecording : startRecording}
          />
          {transcription && <Text style={{ marginTop: 16 }}>{transcription}</Text>}
        </View>
      )
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#transcribe-buffer-from-microphone)

Transcribe Buffer from Microphone

Copy

Ask AI

    // If you have raw audio samples (e.g., from a real-time stream)
    const audioSamples = new Float32Array(rawSamples)
    
    const result = await RunAnywhere.transcribeBuffer(audioSamples, {
      language: 'en',
    })
    
    console.log(result.text)
    

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#supported-audio-formats)

Supported Audio Formats
---------------------------------------------------------------------------------------------------------------

| Format | Extension | Supported |
| --- | --- | --- |
| WAV | `.wav` |     |
| MP3 | `.mp3` |     |
| M4A | `.m4a` |     |
| FLAC | `.flac` |     |
| Raw PCM | \-  | (via buffer methods) |

For best results, use 16kHz mono WAV files. The SDK automatically handles audio conversion, but native formats are faster.

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#available-models)

Available Models
-------------------------------------------------------------------------------------------------

| Model | Size | Languages | Quality | Speed |
| --- | --- | --- | --- | --- |
| whisper-tiny | ~75MB | English | Good | Very Fast |
| whisper-tiny.en | ~75MB | English | Better | Very Fast |
| whisper-base | ~150MB | Multi | Better | Fast |
| whisper-small | ~500MB | Multi | Great | Medium |

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#error-handling)

Error Handling
---------------------------------------------------------------------------------------------

Copy

Ask AI

    import { isSDKError, SDKErrorCode } from '@runanywhere/core'
    
    try {
      const result = await RunAnywhere.transcribe(audioBase64)
    } catch (error) {
      if (isSDKError(error)) {
        switch (error.code) {
          case SDKErrorCode.notInitialized:
            console.error('SDK not initialized')
            break
          case SDKErrorCode.modelNotLoaded:
            console.error('Load an STT model first')
            break
          case SDKErrorCode.sttFailed:
            console.error('Transcription failed:', error.message)
            break
        }
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/stt/transcribe#related)

Related
-------------------------------------------------------------------------------

[STT Streaming\
-------------\
\
Real-time transcription](https://docs.runanywhere.ai/react-native/stt/stream)
[STT Options\
-----------\
\
Advanced configuration](https://docs.runanywhere.ai/react-native/stt/options)
[VAD\
---\
\
Voice Activity Detection](https://docs.runanywhere.ai/react-native/vad)
[Voice Agent\
-----------\
\
Complete voice pipeline](https://docs.runanywhere.ai/react-native/voice-agent)

[System Prompts](https://docs.runanywhere.ai/react-native/llm/system-prompts)
[Streaming STT](https://docs.runanywhere.ai/react-native/stt/stream)

⌘I

---

# Voices - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/tts/voices#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

Voices

On this page

*   [Available Voice Types](https://docs.runanywhere.ai/swift/tts/voices#available-voice-types)
    
*   [Neural Voices (ONNX/Piper)](https://docs.runanywhere.ai/swift/tts/voices#neural-voices-onnx%2Fpiper)
    
*   [System Voices](https://docs.runanywhere.ai/swift/tts/voices#system-voices)
    
*   [Loading Voices](https://docs.runanywhere.ai/swift/tts/voices#loading-voices)
    
*   [Load a Neural Voice](https://docs.runanywhere.ai/swift/tts/voices#load-a-neural-voice)
    
*   [List Available Voices](https://docs.runanywhere.ai/swift/tts/voices#list-available-voices)
    
*   [Check Current Voice](https://docs.runanywhere.ai/swift/tts/voices#check-current-voice)
    
*   [Unload Voice](https://docs.runanywhere.ai/swift/tts/voices#unload-voice)
    
*   [Voice Selection](https://docs.runanywhere.ai/swift/tts/voices#voice-selection)
    
*   [By Language](https://docs.runanywhere.ai/swift/tts/voices#by-language)
    
*   [Voice Picker UI](https://docs.runanywhere.ai/swift/tts/voices#voice-picker-ui)
    
*   [Voice Comparison](https://docs.runanywhere.ai/swift/tts/voices#voice-comparison)
    
*   [Neural vs System Voices](https://docs.runanywhere.ai/swift/tts/voices#neural-vs-system-voices)
    
*   [Quality Tiers](https://docs.runanywhere.ai/swift/tts/voices#quality-tiers)
    
*   [Custom Voice Configuration](https://docs.runanywhere.ai/swift/tts/voices#custom-voice-configuration)
    
*   [Per-Voice Settings](https://docs.runanywhere.ai/swift/tts/voices#per-voice-settings)
    
*   [Downloading Voices](https://docs.runanywhere.ai/swift/tts/voices#downloading-voices)
    
*   [Multi-Voice Support](https://docs.runanywhere.ai/swift/tts/voices#multi-voice-support)
    
*   [Best Practices](https://docs.runanywhere.ai/swift/tts/voices#best-practices)
    

The RunAnywhere SDK supports multiple voice sources: neural ONNX voices (Piper) and system voices via AVSpeechSynthesizer.

[​](https://docs.runanywhere.ai/swift/tts/voices#available-voice-types)

Available Voice Types
------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#neural-voices-onnx/piper)

Neural Voices (ONNX/Piper)

High-quality neural voices that run on-device:

| Voice ID | Language | Gender | Quality | Size |
| --- | --- | --- | --- | --- |
| `piper-en-us-amy` | English | Female | High | ~60 MB |
| `piper-en-us-danny` | English | Male | High | ~60 MB |
| `piper-en-gb-alan` | English | Male | High | ~60 MB |
| `piper-de-thorsten` | German | Male | High | ~60 MB |
| `piper-es-mls` | Spanish | Mixed | High | ~60 MB |
| `piper-fr-mls` | French | Mixed | High | ~60 MB |

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#system-voices)

System Voices

Built-in iOS/macOS voices via AVSpeechSynthesizer:

| Voice ID | Platform |
| --- | --- |
| `com.apple.ttsbundle.siri_female_en-US_compact` | iOS/macOS |
| `com.apple.ttsbundle.siri_male_en-US_compact` | iOS/macOS |
| `com.apple.ttsbundle.Samantha-compact` | iOS/macOS |
| `com.apple.voice.enhanced.en-US.Samantha` | macOS |

[​](https://docs.runanywhere.ai/swift/tts/voices#loading-voices)

Loading Voices
----------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#load-a-neural-voice)

Load a Neural Voice

Copy

Ask AI

    try await RunAnywhere.loadTTSVoice("piper-en-us-amy")
    

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#list-available-voices)

List Available Voices

Copy

Ask AI

    let voices = await RunAnywhere.availableTTSVoices
    for voice in voices {
        print(voice)
    }
    

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#check-current-voice)

Check Current Voice

Copy

Ask AI

    if let currentVoice = await RunAnywhere.currentTTSVoiceId {
        print("Current voice: \(currentVoice)")
    }
    

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#unload-voice)

Unload Voice

Copy

Ask AI

    try await RunAnywhere.unloadTTSVoice()
    

[​](https://docs.runanywhere.ai/swift/tts/voices#voice-selection)

Voice Selection
------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#by-language)

By Language

Copy

Ask AI

    func loadVoiceForLanguage(_ language: String) async throws {
        let voiceMap: [String: String] = [\
            "en": "piper-en-us-amy",\
            "de": "piper-de-thorsten",\
            "es": "piper-es-mls",\
            "fr": "piper-fr-mls"\
        ]
    
        if let voiceId = voiceMap[language] {
            try await RunAnywhere.loadTTSVoice(voiceId)
        } else {
            // Fall back to system voice
            try await RunAnywhere.loadTTSVoice("com.apple.ttsbundle.siri_female_en-US_compact")
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#voice-picker-ui)

Voice Picker UI

Copy

Ask AI

    struct VoicePickerView: View {
        @State private var selectedVoice = "piper-en-us-amy"
        @State private var availableVoices: [String] = []
        @State private var isLoading = false
    
        let voiceNames: [String: String] = [\
            "piper-en-us-amy": "Amy (English, Female)",\
            "piper-en-us-danny": "Danny (English, Male)",\
            "piper-en-gb-alan": "Alan (British, Male)",\
            "piper-de-thorsten": "Thorsten (German, Male)",\
            "piper-es-mls": "Spanish Neural",\
            "piper-fr-mls": "French Neural"\
        ]
    
        var body: some View {
            VStack(spacing: 20) {
                Text("Select Voice")
                    .font(.headline)
    
                Picker("Voice", selection: $selectedVoice) {
                    ForEach(availableVoices, id: \.self) { voice in
                        Text(voiceNames[voice] ?? voice)
                            .tag(voice)
                    }
                }
                .pickerStyle(.wheel)
    
                Button(action: loadSelectedVoice) {
                    HStack {
                        if isLoading {
                            ProgressView()
                                .scaleEffect(0.8)
                        }
                        Text(isLoading ? "Loading..." : "Apply")
                    }
                    .frame(maxWidth: .infinity)
                    .padding()
                    .background(Color.blue)
                    .foregroundColor(.white)
                    .cornerRadius(10)
                }
                .disabled(isLoading)
    
                Button("Preview") {
                    Task {
                        try? await RunAnywhere.speak("Hello! This is a preview of the selected voice.")
                    }
                }
                .disabled(isLoading)
            }
            .padding()
            .task {
                availableVoices = await RunAnywhere.availableTTSVoices
            }
        }
    
        func loadSelectedVoice() {
            isLoading = true
    
            Task {
                do {
                    try await RunAnywhere.loadTTSVoice(selectedVoice)
                } catch {
                    print("Failed to load voice: \(error)")
                }
    
                await MainActor.run {
                    isLoading = false
                }
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/tts/voices#voice-comparison)

Voice Comparison
--------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#neural-vs-system-voices)

Neural vs System Voices

| Feature | Neural (Piper) | System (AVSpeech) |
| --- | --- | --- |
| Quality | Higher | Good |
| Latency | Slightly higher | Lower |
| Offline | Yes | Yes |
| Size | ~60 MB each | Built-in |
| Customization | Rate, pitch | Rate, pitch, volume |
| Languages | Limited selection | Many languages |

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#quality-tiers)

Quality Tiers

Copy

Ask AI

    enum VoiceQuality {
        case high      // Neural voices
        case standard  // Enhanced system voices
        case compact   // Compact system voices
    }
    
    func recommendVoice(for quality: VoiceQuality, language: String) -> String {
        switch quality {
        case .high:
            return "piper-\(language)-*"  // Neural voice
        case .standard:
            return "com.apple.voice.enhanced.\(language)-*"
        case .compact:
            return "com.apple.ttsbundle.*-compact"
        }
    }
    

[​](https://docs.runanywhere.ai/swift/tts/voices#custom-voice-configuration)

Custom Voice Configuration
----------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/tts/voices#per-voice-settings)

Per-Voice Settings

Copy

Ask AI

    struct VoiceSettings {
        let voiceId: String
        var rate: Float = 1.0
        var pitch: Float = 1.0
        var volume: Float = 1.0
    }
    
    class VoiceManager {
        private var settings: [String: VoiceSettings] = [\
            "piper-en-us-amy": VoiceSettings(voiceId: "piper-en-us-amy", rate: 0.95, pitch: 1.0),\
            "piper-en-us-danny": VoiceSettings(voiceId: "piper-en-us-danny", rate: 1.0, pitch: 0.95)\
        ]
    
        func speak(_ text: String) async throws {
            let currentVoice = await RunAnywhere.currentTTSVoiceId ?? "piper-en-us-amy"
            let voiceSettings = settings[currentVoice] ?? VoiceSettings(voiceId: currentVoice)
    
            try await RunAnywhere.speak(
                text,
                options: TTSOptions(
                    rate: voiceSettings.rate,
                    pitch: voiceSettings.pitch,
                    volume: voiceSettings.volume
                )
            )
        }
    }
    

[​](https://docs.runanywhere.ai/swift/tts/voices#downloading-voices)

Downloading Voices
------------------------------------------------------------------------------------------

Neural voices need to be downloaded before first use:

Copy

Ask AI

    // Check if voice is downloaded
    let models = try await RunAnywhere.availableModels()
    let voiceModel = models.first { $0.id == "piper-en-us-amy" }
    
    if let model = voiceModel, !model.isDownloaded {
        // Download the voice
        let task = try await Download.shared.downloadModel(model)
    
        for await progress in task.progress {
            print("Downloading: \(Int(progress.overallProgress * 100))%")
        }
    }
    
    // Now load the voice
    try await RunAnywhere.loadTTSVoice("piper-en-us-amy")
    

[​](https://docs.runanywhere.ai/swift/tts/voices#multi-voice-support)

Multi-Voice Support
--------------------------------------------------------------------------------------------

Switch between voices for different speakers:

Copy

Ask AI

    class MultiVoiceNarrator {
        let voiceAssignments: [String: String] = [\
            "narrator": "piper-en-us-amy",\
            "character1": "piper-en-us-danny",\
            "character2": "piper-en-gb-alan"\
        ]
    
        func speak(as role: String, text: String) async throws {
            guard let voiceId = voiceAssignments[role] else { return }
    
            let currentVoice = await RunAnywhere.currentTTSVoiceId
    
            // Load voice if different
            if currentVoice != voiceId {
                try await RunAnywhere.loadTTSVoice(voiceId)
            }
    
            try await RunAnywhere.speak(text)
        }
    }
    
    // Usage
    let narrator = MultiVoiceNarrator()
    try await narrator.speak(as: "narrator", text: "Once upon a time...")
    try await narrator.speak(as: "character1", text: "Hello, friend!")
    try await narrator.speak(as: "character2", text: "Greetings!")
    

[​](https://docs.runanywhere.ai/swift/tts/voices#best-practices)

Best Practices
----------------------------------------------------------------------------------

Preload voices

Download and load voices at app startup or in onboarding to avoid delays.

Cache voice selection

Save user’s preferred voice to UserDefaults and restore on app launch.

Fallback strategy

Always have a fallback to system voices if neural voices fail to load.

[synthesize()\
------------\
\
Basic synthesis →](https://docs.runanywhere.ai/swift/tts/synthesize)
[Streaming TTS\
-------------\
\
Stream long text →](https://docs.runanywhere.ai/swift/tts/stream)

[Streaming TTS](https://docs.runanywhere.ai/swift/tts/stream)
[Voice Activity Detection](https://docs.runanywhere.ai/swift/vad)

⌘I

---

# Voice Activity Detection - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/vad#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Swift SDK

Voice Activity Detection

On this page

*   [Overview](https://docs.runanywhere.ai/swift/vad#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/swift/vad#basic-usage)
    
*   [Setup](https://docs.runanywhere.ai/swift/vad#setup)
    
*   [VADConfiguration](https://docs.runanywhere.ai/swift/vad#vadconfiguration)
    
*   [Threshold Tuning](https://docs.runanywhere.ai/swift/vad#threshold-tuning)
    
*   [Detection Methods](https://docs.runanywhere.ai/swift/vad#detection-methods)
    
*   [From AVAudioPCMBuffer](https://docs.runanywhere.ai/swift/vad#from-avaudiopcmbuffer)
    
*   [From Float Array](https://docs.runanywhere.ai/swift/vad#from-float-array)
    
*   [Continuous VAD with Callbacks](https://docs.runanywhere.ai/swift/vad#continuous-vad-with-callbacks)
    
*   [Complete Voice Recording Example](https://docs.runanywhere.ai/swift/vad#complete-voice-recording-example)
    
*   [SwiftUI Voice Activity UI](https://docs.runanywhere.ai/swift/vad#swiftui-voice-activity-ui)
    
*   [VAD State Management](https://docs.runanywhere.ai/swift/vad#vad-state-management)
    
*   [Best Practices](https://docs.runanywhere.ai/swift/vad#best-practices)
    
*   [Error Handling](https://docs.runanywhere.ai/swift/vad#error-handling)
    

Voice Activity Detection (VAD) identifies speech in audio streams, enabling hands-free voice interfaces and efficient audio processing by filtering out silence.

[​](https://docs.runanywhere.ai/swift/vad#overview)

Overview
---------------------------------------------------------------

VAD answers the question: “Is someone speaking right now?” This enables:

*   **Wake word detection** – Start listening when speech begins
*   **Audio trimming** – Only process speech segments
*   **Turn-taking** – Know when to respond in conversations
*   **Battery efficiency** – Don’t process silence

[​](https://docs.runanywhere.ai/swift/vad#basic-usage)

Basic Usage
---------------------------------------------------------------------

Copy

Ask AI

    // Initialize VAD
    try await RunAnywhere.initializeVAD()
    
    // Detect speech in audio buffer
    import AVFoundation
    let buffer: AVAudioPCMBuffer = // From microphone
    let isSpeaking = try await RunAnywhere.detectSpeech(in: buffer)
    
    if isSpeaking {
        print("Speech detected!")
    }
    

[​](https://docs.runanywhere.ai/swift/vad#setup)

Setup
---------------------------------------------------------

Copy

Ask AI

    import RunAnywhere
    import ONNXRuntime
    
    // Register ONNX module (at app launch)
    @MainActor
    func setup() {
        ONNX.register()
    }
    
    // Initialize VAD with default configuration
    try await RunAnywhere.initializeVAD()
    
    // Or with custom configuration
    let config = VADConfiguration(
        sampleRate: 16000,
        frameLength: 0.032,     // 32ms frames
        energyThreshold: 0.5    // Sensitivity (0-1)
    )
    try await RunAnywhere.initializeVAD(config)
    

[​](https://docs.runanywhere.ai/swift/vad#vadconfiguration)

VADConfiguration
-------------------------------------------------------------------------------

Copy

Ask AI

    public struct VADConfiguration: Sendable {
        public let sampleRate: Int          // Audio sample rate (default: 16000)
        public let frameLength: Double      // Frame length in seconds (default: 0.032)
        public let energyThreshold: Double  // Detection threshold 0-1 (default: 0.5)
    
        public init(
            sampleRate: Int = 16000,
            frameLength: Double = 0.032,
            energyThreshold: Double = 0.5
        )
    }
    

### 

[​](https://docs.runanywhere.ai/swift/vad#threshold-tuning)

Threshold Tuning

| Threshold | Behavior | Use Case |
| --- | --- | --- |
| 0.2-0.4 | Very sensitive, more false positives | Quiet environments |
| 0.4-0.6 | Balanced (recommended) | Normal rooms |
| 0.6-0.8 | Less sensitive, fewer false positives | Noisy environments |

[​](https://docs.runanywhere.ai/swift/vad#detection-methods)

Detection Methods
---------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/vad#from-avaudiopcmbuffer)

From AVAudioPCMBuffer

Copy

Ask AI

    import AVFoundation
    
    let buffer: AVAudioPCMBuffer = // From audio input
    let isSpeaking = try await RunAnywhere.detectSpeech(in: buffer)
    

### 

[​](https://docs.runanywhere.ai/swift/vad#from-float-array)

From Float Array

Copy

Ask AI

    let samples: [Float] = // Raw audio samples
    let isSpeaking = try await RunAnywhere.detectSpeech(in: samples)
    

[​](https://docs.runanywhere.ai/swift/vad#continuous-vad-with-callbacks)

Continuous VAD with Callbacks
---------------------------------------------------------------------------------------------------------

For real-time applications, use the callback-based API:

Copy

Ask AI

    // Set speech activity callback
    await RunAnywhere.setVADSpeechActivityCallback { event in
        switch event {
        case .started:
            print("User started speaking")
            startRecording()
        case .ended:
            print("User stopped speaking")
            stopRecording()
        }
    }
    
    // Set audio buffer callback (optional)
    await RunAnywhere.setVADAudioBufferCallback { samples in
        // Process audio samples
        visualizeSamples(samples)
    }
    
    // Start VAD processing
    try await RunAnywhere.startVAD()
    
    // ... later
    try await RunAnywhere.stopVAD()
    

[​](https://docs.runanywhere.ai/swift/vad#complete-voice-recording-example)

Complete Voice Recording Example
---------------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class VoiceRecorder: ObservableObject {
        @Published var isListening = false
        @Published var isSpeaking = false
        @Published var recordedAudio: Data?
    
        private var audioEngine: AVAudioEngine?
        private var audioBuffer = Data()
    
        func startListening() async throws {
            // Initialize VAD
            try await RunAnywhere.initializeVAD(
                VADConfiguration(energyThreshold: 0.5)
            )
    
            // Set up speech callbacks
            await RunAnywhere.setVADSpeechActivityCallback { [weak self] event in
                Task { @MainActor in
                    switch event {
                    case .started:
                        self?.isSpeaking = true
                        self?.audioBuffer = Data()
                    case .ended:
                        self?.isSpeaking = false
                        self?.recordedAudio = self?.audioBuffer
                    }
                }
            }
    
            // Set up audio engine
            audioEngine = AVAudioEngine()
            let inputNode = audioEngine!.inputNode
    
            let format = AVAudioFormat(
                commonFormat: .pcmFormatFloat32,
                sampleRate: 16000,
                channels: 1,
                interleaved: false
            )!
    
            inputNode.installTap(onBus: 0, bufferSize: 4096, format: format) { [weak self] buffer, _ in
                self?.processBuffer(buffer)
            }
    
            try audioEngine!.start()
            try await RunAnywhere.startVAD()
    
            await MainActor.run { isListening = true }
        }
    
        private func processBuffer(_ buffer: AVAudioPCMBuffer) {
            Task {
                // Feed to VAD
                let isSpeech = try? await RunAnywhere.detectSpeech(in: buffer)
    
                // If speaking, accumulate audio
                if isSpeech == true, let channelData = buffer.floatChannelData?[0] {
                    let data = Data(bytes: channelData, count: Int(buffer.frameLength) * MemoryLayout<Float>.size)
                    await MainActor.run {
                        audioBuffer.append(data)
                    }
                }
            }
        }
    
        func stopListening() async {
            audioEngine?.stop()
            audioEngine?.inputNode.removeTap(onBus: 0)
            try? await RunAnywhere.stopVAD()
            await RunAnywhere.cleanupVAD()
    
            await MainActor.run { isListening = false }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/vad#swiftui-voice-activity-ui)

SwiftUI Voice Activity UI
-------------------------------------------------------------------------------------------------

Copy

Ask AI

    struct VoiceActivityView: View {
        @StateObject private var recorder = VoiceRecorder()
    
        var body: some View {
            VStack(spacing: 30) {
                // Visual indicator
                ZStack {
                    Circle()
                        .fill(indicatorColor)
                        .frame(width: 120, height: 120)
                        .scaleEffect(recorder.isSpeaking ? 1.2 : 1.0)
                        .animation(.easeInOut(duration: 0.2), value: recorder.isSpeaking)
    
                    Image(systemName: recorder.isSpeaking ? "waveform" : "mic.fill")
                        .font(.system(size: 40))
                        .foregroundColor(.white)
                }
    
                // Status text
                Text(statusText)
                    .font(.headline)
    
                // Control button
                Button(action: toggleListening) {
                    Text(recorder.isListening ? "Stop" : "Start Listening")
                        .frame(maxWidth: .infinity)
                        .padding()
                        .background(recorder.isListening ? Color.red : Color.blue)
                        .foregroundColor(.white)
                        .cornerRadius(10)
                }
                .padding(.horizontal)
    
                // Recorded audio indicator
                if recorder.recordedAudio != nil {
                    Label("Audio captured!", systemImage: "checkmark.circle.fill")
                        .foregroundColor(.green)
                }
            }
            .padding()
        }
    
        var indicatorColor: Color {
            if !recorder.isListening {
                return .gray
            } else if recorder.isSpeaking {
                return .green
            } else {
                return .blue
            }
        }
    
        var statusText: String {
            if !recorder.isListening {
                return "Tap to start"
            } else if recorder.isSpeaking {
                return "Speaking..."
            } else {
                return "Listening..."
            }
        }
    
        func toggleListening() {
            Task {
                if recorder.isListening {
                    await recorder.stopListening()
                } else {
                    try? await recorder.startListening()
                }
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/vad#vad-state-management)

VAD State Management
---------------------------------------------------------------------------------------

Copy

Ask AI

    // Check if VAD is ready
    let isReady = await RunAnywhere.isVADReady
    
    // Start/stop processing
    try await RunAnywhere.startVAD()
    try await RunAnywhere.stopVAD()
    
    // Cleanup resources
    await RunAnywhere.cleanupVAD()
    

[​](https://docs.runanywhere.ai/swift/vad#best-practices)

Best Practices
---------------------------------------------------------------------------

Adjust threshold for environment

Start with 0.5 and adjust based on testing. Noisy environments need higher thresholds.

Add debouncing

Add a small delay before acting on speech end events to handle brief pauses.

Copy

Ask AI

    var speechEndTimer: Task<Void, Never>?
    
    await RunAnywhere.setVADSpeechActivityCallback { event in
        switch event {
        case .started:
            speechEndTimer?.cancel()
            handleSpeechStart()
        case .ended:
            speechEndTimer = Task {
                try? await Task.sleep(for: .milliseconds(500))
                handleSpeechEnd()
            }
        }
    }

Handle background audio

Configure AVAudioSession appropriately to handle interruptions.

Clean up resources

Always call `cleanupVAD()` when done to free resources.

[​](https://docs.runanywhere.ai/swift/vad#error-handling)

Error Handling
---------------------------------------------------------------------------

Copy

Ask AI

    do {
        try await RunAnywhere.initializeVAD()
        try await RunAnywhere.startVAD()
    } catch let error as SDKError {
        switch error.code {
        case .initializationFailed:
            print("VAD initialization failed")
        case .microphonePermissionDenied:
            print("Microphone access required")
        default:
            print("Error: \(error)")
        }
    }
    

[Voice Agent\
-----------\
\
Build complete voice experiences with VAD + STT + LLM + TTS →](https://docs.runanywhere.ai/swift/voice-agent)

[Voices](https://docs.runanywhere.ai/swift/tts/voices)
[Voice Agent](https://docs.runanywhere.ai/swift/voice-agent)

⌘I

---

# STT Options - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/stt/options#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

STT Options

On this page

*   [STTOptions](https://docs.runanywhere.ai/swift/stt/options#sttoptions)
    
*   [Parameters](https://docs.runanywhere.ai/swift/stt/options#parameters)
    
*   [Language Support](https://docs.runanywhere.ai/swift/stt/options#language-support)
    
*   [Auto-Detection](https://docs.runanywhere.ai/swift/stt/options#auto-detection)
    
*   [Force Language](https://docs.runanywhere.ai/swift/stt/options#force-language)
    
*   [Word Timestamps](https://docs.runanywhere.ai/swift/stt/options#word-timestamps)
    
*   [WordTimestamp Structure](https://docs.runanywhere.ai/swift/stt/options#wordtimestamp-structure)
    
*   [Subtitle Generation Example](https://docs.runanywhere.ai/swift/stt/options#subtitle-generation-example)
    
*   [Voice Activity Detection (VAD)](https://docs.runanywhere.ai/swift/stt/options#voice-activity-detection-vad)
    
*   [Sample Rate](https://docs.runanywhere.ai/swift/stt/options#sample-rate)
    
*   [TranscriptionMetadata](https://docs.runanywhere.ai/swift/stt/options#transcriptionmetadata)
    
*   [Alternative Transcriptions](https://docs.runanywhere.ai/swift/stt/options#alternative-transcriptions)
    
*   [Complete Example](https://docs.runanywhere.ai/swift/stt/options#complete-example)
    

The `STTOptions` struct allows you to customize speech-to-text behavior including language, timestamps, and audio format.

[​](https://docs.runanywhere.ai/swift/stt/options#sttoptions)

STTOptions
---------------------------------------------------------------------------

Copy

Ask AI

    public struct STTOptions: Sendable {
        public let language: String
        public let sampleRate: Int
        public let enableWordTimestamps: Bool
        public let enableVAD: Bool
    
        public init(
            language: String = "en",
            sampleRate: Int = 16000,
            enableWordTimestamps: Bool = false,
            enableVAD: Bool = true
        )
    }
    

[​](https://docs.runanywhere.ai/swift/stt/options#parameters)

Parameters
---------------------------------------------------------------------------

| Parameter | Type | Default | Description |
| --- | --- | --- | --- |
| `language` | `String` | `"en"` | Language code (ISO 639-1) |
| `sampleRate` | `Int` | `16000` | Audio sample rate in Hz |
| `enableWordTimestamps` | `Bool` | `false` | Include per-word timing information |
| `enableVAD` | `Bool` | `true` | Use VAD to filter silent segments |

[​](https://docs.runanywhere.ai/swift/stt/options#language-support)

Language Support
---------------------------------------------------------------------------------------

Whisper models support 99 languages. Common language codes:

| Language | Code | Language | Code |
| --- | --- | --- | --- |
| English | `en` | French | `fr` |
| Spanish | `es` | German | `de` |
| Italian | `it` | Portuguese | `pt` |
| Chinese | `zh` | Japanese | `ja` |
| Korean | `ko` | Russian | `ru` |
| Arabic | `ar` | Hindi | `hi` |

### 

[​](https://docs.runanywhere.ai/swift/stt/options#auto-detection)

Auto-Detection

Leave language unspecified or set to empty string for auto-detection:

Copy

Ask AI

    let options = STTOptions()  // Auto-detect language
    
    let output = try await RunAnywhere.transcribeWithOptions(audioData, options: options)
    print("Detected language: \(output.detectedLanguage ?? "unknown")")
    

### 

[​](https://docs.runanywhere.ai/swift/stt/options#force-language)

Force Language

Copy

Ask AI

    // Force Spanish transcription
    let options = STTOptions(language: "es")
    let output = try await RunAnywhere.transcribeWithOptions(audioData, options: options)
    

[​](https://docs.runanywhere.ai/swift/stt/options#word-timestamps)

Word Timestamps
-------------------------------------------------------------------------------------

Enable word-level timing for subtitles, karaoke, or word highlighting:

Copy

Ask AI

    let options = STTOptions(
        language: "en",
        enableWordTimestamps: true
    )
    
    let output = try await RunAnywhere.transcribeWithOptions(audioData, options: options)
    
    if let timestamps = output.wordTimestamps {
        for word in timestamps {
            print("\(word.word): \(word.startTime)s - \(word.endTime)s")
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/stt/options#wordtimestamp-structure)

WordTimestamp Structure

Copy

Ask AI

    public struct WordTimestamp: Sendable {
        public let word: String       // The word
        public let startTime: Double  // Start time in seconds
        public let endTime: Double    // End time in seconds
        public let confidence: Float? // Word-level confidence
    }
    

### 

[​](https://docs.runanywhere.ai/swift/stt/options#subtitle-generation-example)

Subtitle Generation Example

Copy

Ask AI

    func generateSubtitles(from audioData: Data) async throws -> [Subtitle] {
        let options = STTOptions(
            language: "en",
            enableWordTimestamps: true
        )
    
        let output = try await RunAnywhere.transcribeWithOptions(audioData, options: options)
    
        guard let timestamps = output.wordTimestamps else {
            return [Subtitle(text: output.text, start: 0, end: 5)]
        }
    
        // Group words into subtitle segments (max 7 words per segment)
        var subtitles: [Subtitle] = []
        var currentWords: [WordTimestamp] = []
    
        for word in timestamps {
            currentWords.append(word)
    
            if currentWords.count >= 7 {
                let text = currentWords.map(\.word).joined(separator: " ")
                let start = currentWords.first!.startTime
                let end = currentWords.last!.endTime
                subtitles.append(Subtitle(text: text, start: start, end: end))
                currentWords = []
            }
        }
    
        // Add remaining words
        if !currentWords.isEmpty {
            let text = currentWords.map(\.word).joined(separator: " ")
            let start = currentWords.first!.startTime
            let end = currentWords.last!.endTime
            subtitles.append(Subtitle(text: text, start: start, end: end))
        }
    
        return subtitles
    }
    
    struct Subtitle {
        let text: String
        let start: Double
        let end: Double
    }
    

[​](https://docs.runanywhere.ai/swift/stt/options#voice-activity-detection-vad)

Voice Activity Detection (VAD)
-----------------------------------------------------------------------------------------------------------------

VAD filters out silent segments before transcription:

Copy

Ask AI

    // With VAD (default) - silent parts are skipped
    let withVAD = STTOptions(enableVAD: true)
    
    // Without VAD - entire audio is processed
    let withoutVAD = STTOptions(enableVAD: false)
    

Enable VAD for real-time transcription to reduce processing time and improve accuracy by ignoring silence.

[​](https://docs.runanywhere.ai/swift/stt/options#sample-rate)

Sample Rate
-----------------------------------------------------------------------------

The SDK expects audio at 16kHz by default. If your audio has a different sample rate, specify it:

Copy

Ask AI

    // For 44.1kHz audio (resampling will be applied)
    let options = STTOptions(sampleRate: 44100)
    
    // For 48kHz audio
    let options = STTOptions(sampleRate: 48000)
    

For best results, record audio at 16kHz mono. While the SDK can resample, native 16kHz audio produces better accuracy.

[​](https://docs.runanywhere.ai/swift/stt/options#transcriptionmetadata)

TranscriptionMetadata
-------------------------------------------------------------------------------------------------

Get information about the transcription process:

Copy

Ask AI

    let output = try await RunAnywhere.transcribeWithOptions(audioData, options: options)
    
    if let metadata = output.metadata {
        print("Model: \(metadata.modelId)")
        print("Processing time: \(metadata.processingTime)s")
        print("Audio duration: \(metadata.audioLength)s")
    
        let realTimeFactor = metadata.processingTime / metadata.audioLength
        print("Real-time factor: \(String(format: "%.2f", realTimeFactor))x")
    }
    

[​](https://docs.runanywhere.ai/swift/stt/options#alternative-transcriptions)

Alternative Transcriptions
-----------------------------------------------------------------------------------------------------------

Access alternative interpretations:

Copy

Ask AI

    let output = try await RunAnywhere.transcribeWithOptions(audioData, options: options)
    
    print("Best: \(output.text)")
    
    if let alternatives = output.alternatives {
        for (i, alt) in alternatives.enumerated() {
            print("Alt \(i + 1): \(alt.text) (confidence: \(alt.confidence ?? 0))")
        }
    }
    

[​](https://docs.runanywhere.ai/swift/stt/options#complete-example)

Complete Example
---------------------------------------------------------------------------------------

Copy

Ask AI

    class TranscriptionService {
        func transcribe(
            audioURL: URL,
            language: String? = nil,
            includeTimestamps: Bool = false
        ) async throws -> TranscriptionResult {
            let audioData = try Data(contentsOf: audioURL)
    
            let options = STTOptions(
                language: language ?? "en",
                sampleRate: 16000,
                enableWordTimestamps: includeTimestamps,
                enableVAD: true
            )
    
            let output = try await RunAnywhere.transcribeWithOptions(audioData, options: options)
    
            return TranscriptionResult(
                text: output.text,
                confidence: output.confidence,
                language: output.detectedLanguage ?? language ?? "en",
                timestamps: output.wordTimestamps,
                processingTime: output.metadata?.processingTime
            )
        }
    }
    
    struct TranscriptionResult {
        let text: String
        let confidence: Float?
        let language: String
        let timestamps: [WordTimestamp]?
        let processingTime: TimeInterval?
    }
    

[transcribe()\
------------\
\
Basic transcription →](https://docs.runanywhere.ai/swift/stt/transcribe)
[Streaming STT\
-------------\
\
Real-time transcription →](https://docs.runanywhere.ai/swift/stt/stream)

[Streaming STT](https://docs.runanywhere.ai/swift/stt/stream)
[synthesize()](https://docs.runanywhere.ai/swift/tts/synthesize)

⌘I

---

# Streaming STT - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/stt/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

Streaming STT

On this page

*   [Basic Usage](https://docs.runanywhere.ai/swift/stt/stream#basic-usage)
    
*   [Method Signature](https://docs.runanywhere.ai/swift/stt/stream#method-signature)
    
*   [Parameters](https://docs.runanywhere.ai/swift/stt/stream#parameters)
    
*   [Returns](https://docs.runanywhere.ai/swift/stt/stream#returns)
    
*   [STTTranscriptionResult](https://docs.runanywhere.ai/swift/stt/stream#stttranscriptionresult)
    
*   [Live Transcription Example](https://docs.runanywhere.ai/swift/stt/stream#live-transcription-example)
    
*   [SwiftUI Integration](https://docs.runanywhere.ai/swift/stt/stream#swiftui-integration)
    
*   [Voice Command Detection](https://docs.runanywhere.ai/swift/stt/stream#voice-command-detection)
    
*   [Performance Tips](https://docs.runanywhere.ai/swift/stt/stream#performance-tips)
    
*   [Error Handling](https://docs.runanywhere.ai/swift/stt/stream#error-handling)
    

Stream transcription provides real-time results as audio is being processed, enabling responsive voice interfaces.

[​](https://docs.runanywhere.ai/swift/stt/stream#basic-usage)

Basic Usage
----------------------------------------------------------------------------

Copy

Ask AI

    let output = try await RunAnywhere.transcribeStream(
        audioData: audioData,
        options: STTOptions(language: "en"),
        onPartialResult: { result in
            print("Partial: \(result.transcript)")
        }
    )
    
    print("Final: \(output.text)")
    

[​](https://docs.runanywhere.ai/swift/stt/stream#method-signature)

Method Signature
--------------------------------------------------------------------------------------

Copy

Ask AI

    public static func transcribeStream(
        audioData: Data,
        options: STTOptions = STTOptions(),
        onPartialResult: @escaping (STTTranscriptionResult) -> Void
    ) async throws -> STTOutput
    

### 

[​](https://docs.runanywhere.ai/swift/stt/stream#parameters)

Parameters

| Parameter | Type | Description |
| --- | --- | --- |
| `audioData` | `Data` | Audio data to transcribe |
| `options` | `STTOptions` | Transcription options |
| `onPartialResult` | `(STTTranscriptionResult) -> Void` | Callback for partial results |

### 

[​](https://docs.runanywhere.ai/swift/stt/stream#returns)

Returns

`STTOutput` with the final complete transcription.

[​](https://docs.runanywhere.ai/swift/stt/stream#stttranscriptionresult)

STTTranscriptionResult
--------------------------------------------------------------------------------------------------

Copy

Ask AI

    public struct STTTranscriptionResult: Sendable {
        public let transcript: String           // Current transcription
        public let confidence: Float?           // Confidence score
        public let timestamps: [WordTimestamp]? // Word-level timing
        public let language: String?            // Detected language
        public let alternatives: [STTAlternative]? // Alternative transcriptions
    }
    

[​](https://docs.runanywhere.ai/swift/stt/stream#live-transcription-example)

Live Transcription Example
----------------------------------------------------------------------------------------------------------

Copy

Ask AI

    class LiveTranscriber: ObservableObject {
        @Published var partialText = ""
        @Published var finalText = ""
        @Published var isTranscribing = false
    
        private var audioEngine: AVAudioEngine?
        private var audioBuffer = Data()
    
        func startLiveTranscription() async throws {
            let audioSession = AVAudioSession.sharedInstance()
            try audioSession.setCategory(.playAndRecord, mode: .measurement)
            try audioSession.setActive(true)
    
            audioEngine = AVAudioEngine()
            guard let audioEngine = audioEngine else { return }
    
            let inputNode = audioEngine.inputNode
            let recordingFormat = AVAudioFormat(
                commonFormat: .pcmFormatFloat32,
                sampleRate: 16000,
                channels: 1,
                interleaved: false
            )!
    
            isTranscribing = true
    
            inputNode.installTap(onBus: 0, bufferSize: 4096, format: recordingFormat) { [weak self] buffer, _ in
                self?.processAudioBuffer(buffer)
            }
    
            audioEngine.prepare()
            try audioEngine.start()
        }
    
        private func processAudioBuffer(_ buffer: AVAudioPCMBuffer) {
            // Convert buffer to Data and accumulate
            guard let channelData = buffer.floatChannelData?[0] else { return }
            let frameCount = Int(buffer.frameLength)
    
            let data = Data(bytes: channelData, count: frameCount * MemoryLayout<Float>.size)
            audioBuffer.append(data)
    
            // Process every ~1 second of audio
            if audioBuffer.count >= 16000 * 4 { // 1 second at 16kHz, Float32
                let chunk = audioBuffer
                audioBuffer = Data()
    
                Task {
                    await transcribeChunk(chunk)
                }
            }
        }
    
        private func transcribeChunk(_ data: Data) async {
            do {
                let output = try await RunAnywhere.transcribeStream(
                    audioData: data,
                    options: STTOptions(language: "en"),
                    onPartialResult: { [weak self] result in
                        Task { @MainActor in
                            self?.partialText = result.transcript
                        }
                    }
                )
    
                await MainActor.run {
                    finalText += " " + output.text
                    partialText = ""
                }
            } catch {
                print("Transcription error: \(error)")
            }
        }
    
        func stop() {
            audioEngine?.stop()
            audioEngine?.inputNode.removeTap(onBus: 0)
            audioEngine = nil
            isTranscribing = false
        }
    }
    

[​](https://docs.runanywhere.ai/swift/stt/stream#swiftui-integration)

SwiftUI Integration
--------------------------------------------------------------------------------------------

Copy

Ask AI

    struct LiveTranscriptionView: View {
        @StateObject private var transcriber = LiveTranscriber()
    
        var body: some View {
            VStack(spacing: 20) {
                // Final transcription
                ScrollView {
                    Text(transcriber.finalText)
                        .frame(maxWidth: .infinity, alignment: .leading)
                }
                .frame(height: 200)
                .padding()
                .background(Color.gray.opacity(0.1))
                .cornerRadius(12)
    
                // Partial (in-progress) transcription
                if !transcriber.partialText.isEmpty {
                    Text(transcriber.partialText)
                        .foregroundColor(.secondary)
                        .italic()
                }
    
                // Recording indicator
                HStack {
                    if transcriber.isTranscribing {
                        Circle()
                            .fill(Color.red)
                            .frame(width: 12, height: 12)
                        Text("Listening...")
                    }
                }
    
                // Control buttons
                HStack(spacing: 20) {
                    Button(action: startTranscription) {
                        Label("Start", systemImage: "mic.fill")
                            .padding()
                            .background(Color.green)
                            .foregroundColor(.white)
                            .cornerRadius(10)
                    }
                    .disabled(transcriber.isTranscribing)
    
                    Button(action: stopTranscription) {
                        Label("Stop", systemImage: "stop.fill")
                            .padding()
                            .background(Color.red)
                            .foregroundColor(.white)
                            .cornerRadius(10)
                    }
                    .disabled(!transcriber.isTranscribing)
                }
            }
            .padding()
        }
    
        func startTranscription() {
            Task {
                try await transcriber.startLiveTranscription()
            }
        }
    
        func stopTranscription() {
            transcriber.stop()
        }
    }
    

[​](https://docs.runanywhere.ai/swift/stt/stream#voice-command-detection)

Voice Command Detection
----------------------------------------------------------------------------------------------------

Use streaming STT to detect commands in real-time:

Copy

Ask AI

    class VoiceCommandDetector {
        private let commands = ["play", "pause", "stop", "next", "previous"]
        var onCommandDetected: ((String) -> Void)?
    
        func processAudio(_ data: Data) async {
            do {
                let _ = try await RunAnywhere.transcribeStream(
                    audioData: data,
                    options: STTOptions(language: "en"),
                    onPartialResult: { [weak self] result in
                        self?.checkForCommands(result.transcript)
                    }
                )
            } catch {
                print("Error: \(error)")
            }
        }
    
        private func checkForCommands(_ text: String) {
            let lowercased = text.lowercased()
            for command in commands {
                if lowercased.contains(command) {
                    onCommandDetected?(command)
                    break
                }
            }
        }
    }
    
    // Usage
    let detector = VoiceCommandDetector()
    detector.onCommandDetected = { command in
        switch command {
        case "play":
            player.play()
        case "pause":
            player.pause()
        default:
            break
        }
    }
    

[​](https://docs.runanywhere.ai/swift/stt/stream#performance-tips)

Performance Tips
--------------------------------------------------------------------------------------

Chunk size

Process audio in chunks of 1-2 seconds for best balance of latency and accuracy.

Buffer management

Clear audio buffers after processing to prevent memory growth during long sessions.

Background processing

Transcription runs in background threads automatically. Update UI on MainActor.

[​](https://docs.runanywhere.ai/swift/stt/stream#error-handling)

Error Handling
----------------------------------------------------------------------------------

Copy

Ask AI

    do {
        let output = try await RunAnywhere.transcribeStream(
            audioData: data,
            onPartialResult: { _ in }
        )
    } catch let error as SDKError {
        switch error.code {
        case .modelNotFound:
            print("Load STT model first")
        case .emptyAudioBuffer:
            print("No audio data")
        case .processingFailed:
            print("Streaming failed: \(error.message)")
        default:
            print("Error: \(error)")
        }
    }
    

[Basic Transcription\
-------------------\
\
Non-streaming transcription →](https://docs.runanywhere.ai/swift/stt/transcribe)
[STT Options\
-----------\
\
Configure transcription →](https://docs.runanywhere.ai/swift/stt/options)

[transcribe()](https://docs.runanywhere.ai/swift/stt/transcribe)
[STT Options](https://docs.runanywhere.ai/swift/stt/options)

⌘I

---

# transcribe() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/stt/transcribe#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

transcribe()

On this page

*   [Basic Usage](https://docs.runanywhere.ai/swift/stt/transcribe#basic-usage)
    
*   [Setup](https://docs.runanywhere.ai/swift/stt/transcribe#setup)
    
*   [Method Signatures](https://docs.runanywhere.ai/swift/stt/transcribe#method-signatures)
    
*   [Simple Transcription](https://docs.runanywhere.ai/swift/stt/transcribe#simple-transcription)
    
*   [Transcription with Options](https://docs.runanywhere.ai/swift/stt/transcribe#transcription-with-options)
    
*   [Buffer Transcription](https://docs.runanywhere.ai/swift/stt/transcribe#buffer-transcription)
    
*   [Audio Requirements](https://docs.runanywhere.ai/swift/stt/transcribe#audio-requirements)
    
*   [STTOutput](https://docs.runanywhere.ai/swift/stt/transcribe#sttoutput)
    
*   [Examples](https://docs.runanywhere.ai/swift/stt/transcribe#examples)
    
*   [Recording and Transcribing](https://docs.runanywhere.ai/swift/stt/transcribe#recording-and-transcribing)
    
*   [With Timestamps](https://docs.runanywhere.ai/swift/stt/transcribe#with-timestamps)
    
*   [Multi-Language Support](https://docs.runanywhere.ai/swift/stt/transcribe#multi-language-support)
    
*   [SwiftUI Voice Input](https://docs.runanywhere.ai/swift/stt/transcribe#swiftui-voice-input)
    
*   [Model Management](https://docs.runanywhere.ai/swift/stt/transcribe#model-management)
    
*   [Available Models](https://docs.runanywhere.ai/swift/stt/transcribe#available-models)
    
*   [Error Handling](https://docs.runanywhere.ai/swift/stt/transcribe#error-handling)
    

The `transcribe()` method converts audio data to text using on-device speech recognition models like Whisper.

[​](https://docs.runanywhere.ai/swift/stt/transcribe#basic-usage)

Basic Usage
--------------------------------------------------------------------------------

Copy

Ask AI

    // Simple transcription
    let text = try await RunAnywhere.transcribe(audioData)
    print("You said: \(text)")
    

[​](https://docs.runanywhere.ai/swift/stt/transcribe#setup)

Setup
--------------------------------------------------------------------

Before transcribing, register the ONNX module and load an STT model:

Copy

Ask AI

    import RunAnywhere
    import ONNXRuntime
    
    // Register ONNX module (at app launch)
    @MainActor
    func setup() {
        ONNX.register()
    }
    
    // Load STT model
    try await RunAnywhere.loadSTTModel("whisper-base-onnx")
    

[​](https://docs.runanywhere.ai/swift/stt/transcribe#method-signatures)

Method Signatures
--------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/stt/transcribe#simple-transcription)

Simple Transcription

Copy

Ask AI

    public static func transcribe(_ audioData: Data) async throws -> String
    

Returns just the transcribed text.

### 

[​](https://docs.runanywhere.ai/swift/stt/transcribe#transcription-with-options)

Transcription with Options

Copy

Ask AI

    public static func transcribeWithOptions(
        _ audioData: Data,
        options: STTOptions
    ) async throws -> STTOutput
    

Returns detailed output including confidence and timestamps.

### 

[​](https://docs.runanywhere.ai/swift/stt/transcribe#buffer-transcription)

Buffer Transcription

Copy

Ask AI

    public static func transcribeBuffer(
        _ buffer: AVAudioPCMBuffer,
        language: String? = nil
    ) async throws -> STTOutput
    

Transcribe directly from an `AVAudioPCMBuffer`.

[​](https://docs.runanywhere.ai/swift/stt/transcribe#audio-requirements)

Audio Requirements
----------------------------------------------------------------------------------------------

| Property | Requirement |
| --- | --- |
| Sample Rate | 16,000 Hz (recommended) |
| Channels | Mono (1 channel) |
| Format | Float32 or Int16 PCM |
| Duration | Up to 30 seconds per call (Whisper limitation) |

[​](https://docs.runanywhere.ai/swift/stt/transcribe#sttoutput)

STTOutput
----------------------------------------------------------------------------

Copy

Ask AI

    public struct STTOutput: Sendable {
        public let text: String                       // Transcribed text
        public let confidence: Float?                 // Confidence score (0-1)
        public let wordTimestamps: [WordTimestamp]?   // Per-word timing
        public let detectedLanguage: String?          // Detected language code
        public let alternatives: [STTAlternative]?    // Alternative transcriptions
        public let metadata: TranscriptionMetadata?   // Processing info
    }
    

[​](https://docs.runanywhere.ai/swift/stt/transcribe#examples)

Examples
--------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/stt/transcribe#recording-and-transcribing)

Recording and Transcribing

Copy

Ask AI

    import AVFoundation
    
    class AudioRecorder: ObservableObject {
        private var audioRecorder: AVAudioRecorder?
        private let audioSession = AVAudioSession.sharedInstance()
    
        func startRecording() async throws {
            try audioSession.setCategory(.playAndRecord, mode: .default)
            try audioSession.setActive(true)
    
            let url = FileManager.default.temporaryDirectory
                .appendingPathComponent("recording.wav")
    
            let settings: [String: Any] = [\
                AVFormatIDKey: Int(kAudioFormatLinearPCM),\
                AVSampleRateKey: 16000,\
                AVNumberOfChannelsKey: 1,\
                AVLinearPCMBitDepthKey: 16,\
                AVLinearPCMIsFloatKey: false\
            ]
    
            audioRecorder = try AVAudioRecorder(url: url, settings: settings)
            audioRecorder?.record()
        }
    
        func stopAndTranscribe() async throws -> String {
            guard let recorder = audioRecorder else { return "" }
    
            recorder.stop()
            let audioData = try Data(contentsOf: recorder.url)
    
            return try await RunAnywhere.transcribe(audioData)
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/stt/transcribe#with-timestamps)

With Timestamps

Copy

Ask AI

    let output = try await RunAnywhere.transcribeWithOptions(
        audioData,
        options: STTOptions(
            language: "en",
            enableWordTimestamps: true
        )
    )
    
    print("Full text: \(output.text)")
    
    if let timestamps = output.wordTimestamps {
        for word in timestamps {
            print("\(word.word): \(word.startTime)s - \(word.endTime)s")
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/stt/transcribe#multi-language-support)

Multi-Language Support

Copy

Ask AI

    // Auto-detect language
    let output = try await RunAnywhere.transcribeWithOptions(
        audioData,
        options: STTOptions()
    )
    print("Detected language: \(output.detectedLanguage ?? "unknown")")
    
    // Force specific language
    let spanishOutput = try await RunAnywhere.transcribeWithOptions(
        audioData,
        options: STTOptions(language: "es")
    )
    

### 

[​](https://docs.runanywhere.ai/swift/stt/transcribe#swiftui-voice-input)

SwiftUI Voice Input

Copy

Ask AI

    struct VoiceInputView: View {
        @StateObject private var recorder = AudioRecorder()
        @State private var transcription = ""
        @State private var isRecording = false
        @State private var isProcessing = false
    
        var body: some View {
            VStack(spacing: 20) {
                // Transcription display
                Text(transcription)
                    .frame(maxWidth: .infinity, minHeight: 100, alignment: .topLeading)
                    .padding()
                    .background(Color.gray.opacity(0.1))
                    .cornerRadius(12)
    
                // Record button
                Button(action: toggleRecording) {
                    ZStack {
                        Circle()
                            .fill(isRecording ? Color.red : Color.blue)
                            .frame(width: 80, height: 80)
    
                        if isProcessing {
                            ProgressView()
                                .tint(.white)
                        } else {
                            Image(systemName: isRecording ? "stop.fill" : "mic.fill")
                                .font(.title)
                                .foregroundColor(.white)
                        }
                    }
                }
                .disabled(isProcessing)
    
                Text(isRecording ? "Tap to stop" : "Tap to record")
                    .font(.caption)
                    .foregroundColor(.secondary)
            }
            .padding()
        }
    
        func toggleRecording() {
            if isRecording {
                stopRecording()
            } else {
                startRecording()
            }
        }
    
        func startRecording() {
            Task {
                do {
                    try await recorder.startRecording()
                    isRecording = true
                } catch {
                    transcription = "Failed to start recording: \(error)"
                }
            }
        }
    
        func stopRecording() {
            isRecording = false
            isProcessing = true
    
            Task {
                do {
                    let text = try await recorder.stopAndTranscribe()
                    await MainActor.run {
                        transcription = text
                        isProcessing = false
                    }
                } catch {
                    await MainActor.run {
                        transcription = "Transcription failed: \(error)"
                        isProcessing = false
                    }
                }
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/stt/transcribe#model-management)

Model Management
------------------------------------------------------------------------------------------

Copy

Ask AI

    // Load a model
    try await RunAnywhere.loadSTTModel("whisper-base-onnx")
    
    // Check if loaded
    let isLoaded = await RunAnywhere.isSTTModelLoaded
    
    // Get current model
    let model = await RunAnywhere.currentSTTModel
    
    // Unload when done
    try await RunAnywhere.unloadSTTModel()
    

[​](https://docs.runanywhere.ai/swift/stt/transcribe#available-models)

Available Models
------------------------------------------------------------------------------------------

| Model ID | Size | Quality | Speed |
| --- | --- | --- | --- |
| `whisper-tiny-onnx` | ~40MB | Good | Fastest |
| `whisper-base-onnx` | ~150MB | Better | Fast |
| `whisper-small-onnx` | ~500MB | Best | Slower |

[​](https://docs.runanywhere.ai/swift/stt/transcribe#error-handling)

Error Handling
--------------------------------------------------------------------------------------

Copy

Ask AI

    do {
        let text = try await RunAnywhere.transcribe(audioData)
        print(text)
    } catch let error as SDKError {
        switch error.code {
        case .modelNotFound:
            print("Load an STT model first")
        case .emptyAudioBuffer:
            print("Audio data is empty")
        case .processingFailed:
            print("Transcription failed: \(error.message)")
        default:
            print("Error: \(error.localizedDescription)")
        }
    }
    

[Streaming STT\
-------------\
\
Real-time transcription →](https://docs.runanywhere.ai/swift/stt/stream)
[STT Options\
-----------\
\
Configure transcription →](https://docs.runanywhere.ai/swift/stt/options)

[System Prompts](https://docs.runanywhere.ai/swift/llm/system-prompts)
[Streaming STT](https://docs.runanywhere.ai/swift/stt/stream)

⌘I

---

# Streaming TTS - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/tts/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

Streaming TTS

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/tts/stream#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/react-native/tts/stream#basic-usage)
    
*   [API Reference](https://docs.runanywhere.ai/react-native/tts/stream#api-reference)
    
*   [synthesizeStream](https://docs.runanywhere.ai/react-native/tts/stream#synthesizestream)
    
*   [TTSOutput (Chunk)](https://docs.runanywhere.ai/react-native/tts/stream#ttsoutput-chunk)
    
*   [Examples](https://docs.runanywhere.ai/react-native/tts/stream#examples)
    
*   [Streaming Playback](https://docs.runanywhere.ai/react-native/tts/stream#streaming-playback)
    
*   [Audio Queue Manager](https://docs.runanywhere.ai/react-native/tts/stream#audio-queue-manager)
    
*   [Custom Hook](https://docs.runanywhere.ai/react-native/tts/stream#custom-hook)
    
*   [With Progress Tracking](https://docs.runanywhere.ai/react-native/tts/stream#with-progress-tracking)
    
*   [Performance Considerations](https://docs.runanywhere.ai/react-native/tts/stream#performance-considerations)
    
*   [When to Use Streaming](https://docs.runanywhere.ai/react-native/tts/stream#when-to-use-streaming)
    
*   [Buffer Size](https://docs.runanywhere.ai/react-native/tts/stream#buffer-size)
    
*   [Error Handling](https://docs.runanywhere.ai/react-native/tts/stream#error-handling)
    
*   [Related](https://docs.runanywhere.ai/react-native/tts/stream#related)
    

[​](https://docs.runanywhere.ai/react-native/tts/stream#overview)

Overview
-----------------------------------------------------------------------------

TTS streaming allows you to play audio as it’s being synthesized, providing faster time-to-first-audio for long text. This is especially useful for reading long articles or generating AI responses.

[​](https://docs.runanywhere.ai/react-native/tts/stream#basic-usage)

Basic Usage
-----------------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    // Stream synthesis with callback
    await RunAnywhere.synthesizeStream(
      'This is a longer piece of text that will be synthesized and streamed in chunks for faster playback.',
      { rate: 1.0 },
      (chunk) => {
        // Each chunk contains partial audio
        console.log('Got chunk:', chunk.numSamples, 'samples')
        playChunk(chunk.audio, chunk.sampleRate)
      }
    )
    

[​](https://docs.runanywhere.ai/react-native/tts/stream#api-reference)

API Reference
---------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/tts/stream#synthesizestream)

`synthesizeStream`

Synthesize text with streaming audio chunks.

Copy

Ask AI

    await RunAnywhere.synthesizeStream(
      text: string,
      options?: TTSConfiguration,
      callback?: (chunk: TTSOutput) => void
    ): Promise<TTSResult>
    

**Parameters:**

| Parameter | Type | Description |
| --- | --- | --- |
| `text` | `string` | Text to synthesize |
| `options` | `TTSConfiguration` | Voice configuration |
| `callback` | `(chunk: TTSOutput) => void` | Called for each audio chunk |

### 

[​](https://docs.runanywhere.ai/react-native/tts/stream#ttsoutput-chunk)

TTSOutput (Chunk)

Copy

Ask AI

    interface TTSOutput {
      /** Base64-encoded audio chunk (float32 PCM) */
      audio: string
    
      /** Sample rate in Hz */
      sampleRate: number
    
      /** Number of samples in this chunk */
      numSamples: number
    
      /** Duration of this chunk in seconds */
      duration: number
    
      /** Whether this is the final chunk */
      isFinal: boolean
    }
    

[​](https://docs.runanywhere.ai/react-native/tts/stream#examples)

Examples
-----------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/tts/stream#streaming-playback)

Streaming Playback

StreamingTTS.tsx

Copy

Ask AI

    import React, { useState, useCallback, useRef } from 'react'
    import { View, Button, TextInput, Text, ActivityIndicator } from 'react-native'
    import { RunAnywhere, TTSOutput } from '@runanywhere/core'
    
    export function StreamingTTS() {
      const [text, setText] = useState('')
      const [isStreaming, setIsStreaming] = useState(false)
      const [progress, setProgress] = useState(0)
      const audioQueueRef = useRef<TTSOutput[]>([])
    
      const handleStream = useCallback(async () => {
        if (!text.trim()) return
    
        setIsStreaming(true)
        setProgress(0)
        audioQueueRef.current = []
    
        let totalDuration = 0
        let processedDuration = 0
    
        try {
          const result = await RunAnywhere.synthesizeStream(text, { rate: 1.0 }, (chunk) => {
            // Queue chunk for playback
            audioQueueRef.current.push(chunk)
    
            // Update progress
            processedDuration += chunk.duration
            if (totalDuration > 0) {
              setProgress(processedDuration / totalDuration)
            }
    
            // Play chunk immediately
            playAudioChunk(chunk)
          })
    
          totalDuration = result.duration
          console.log('Total duration:', result.duration, 'seconds')
        } catch (error) {
          console.error('Streaming failed:', error)
        } finally {
          setIsStreaming(false)
        }
      }, [text])
    
      return (
        <View style={{ padding: 16 }}>
          <TextInput
            value={text}
            onChangeText={setText}
            placeholder="Enter text to stream..."
            multiline
            numberOfLines={4}
            style={{ borderWidth: 1, padding: 12 }}
          />
    
          <Button
            title={isStreaming ? 'Streaming...' : 'Stream'}
            onPress={handleStream}
            disabled={isStreaming || !text.trim()}
          />
    
          {isStreaming && (
            <View style={{ marginTop: 16, flexDirection: 'row', alignItems: 'center' }}>
              <ActivityIndicator size="small" />
              <Text style={{ marginLeft: 8 }}>Progress: {(progress * 100).toFixed(0)}%</Text>
            </View>
          )}
        </View>
      )
    }
    
    // Helper to play audio chunk
    function playAudioChunk(chunk: TTSOutput) {
      // Implementation depends on your audio library
      console.log('Playing chunk:', chunk.duration, 'seconds')
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/tts/stream#audio-queue-manager)

Audio Queue Manager

For smooth playback, manage an audio queue using the SDK’s `Audio` utility:

Copy

Ask AI

    import { RunAnywhere, TTSOutput } from '@runanywhere/core'
    
    class AudioQueuePlayer {
      private queue: TTSOutput[] = []
      private isPlaying = false
    
      enqueue(chunk: TTSOutput) {
        this.queue.push(chunk)
        if (!this.isPlaying) {
          this.playNext()
        }
      }
    
      private async playNext() {
        if (this.queue.length === 0) {
          this.isPlaying = false
          return
        }
    
        this.isPlaying = true
        const chunk = this.queue.shift()!
    
        // Convert float32 PCM to WAV and play
        const wavBase64 = RunAnywhere.Audio.createWavFromPCMFloat32(chunk.audio, chunk.sampleRate)
        await RunAnywhere.Audio.playAudio(wavBase64)
    
        // Play next chunk
        this.playNext()
      }
    
      async stop() {
        this.queue = []
        this.isPlaying = false
        await RunAnywhere.Audio.stopPlayback()
      }
    }
    
    // Usage
    const player = new AudioQueuePlayer()
    
    await RunAnywhere.synthesizeStream(longText, { rate: 1.0 }, (chunk) => player.enqueue(chunk))
    

`AudioContext` / Web Audio API does **not** work in React Native. Always use `RunAnywhere.Audio.createWavFromPCMFloat32()` to convert TTS output before playback.

### 

[​](https://docs.runanywhere.ai/react-native/tts/stream#custom-hook)

Custom Hook

useTTSStream.ts

Copy

Ask AI

    import { useState, useCallback, useRef } from 'react'
    import { RunAnywhere, TTSConfiguration, TTSOutput, TTSResult } from '@runanywhere/core'
    
    interface StreamState {
      isStreaming: boolean
      chunks: TTSOutput[]
      result: TTSResult | null
      error: Error | null
    }
    
    export function useTTSStream() {
      const [state, setState] = useState<StreamState>({
        isStreaming: false,
        chunks: [],
        result: null,
        error: null,
      })
    
      const stream = useCallback(
        async (text: string, options?: TTSConfiguration, onChunk?: (chunk: TTSOutput) => void) => {
          setState({ isStreaming: true, chunks: [], result: null, error: null })
    
          try {
            const result = await RunAnywhere.synthesizeStream(text, options, (chunk) => {
              setState((s) => ({ ...s, chunks: [...s.chunks, chunk] }))
              onChunk?.(chunk)
            })
    
            setState((s) => ({ ...s, isStreaming: false, result }))
            return result
          } catch (error) {
            const e = error instanceof Error ? error : new Error('Stream failed')
            setState((s) => ({ ...s, isStreaming: false, error: e }))
            throw e
          }
        },
        []
      )
    
      const reset = useCallback(() => {
        setState({ isStreaming: false, chunks: [], result: null, error: null })
      }, [])
    
      return { ...state, stream, reset }
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/tts/stream#with-progress-tracking)

With Progress Tracking

Copy

Ask AI

    async function synthesizeWithProgress(
      text: string,
      onProgress: (percent: number) => void
    ): Promise<TTSResult> {
      let totalChunks = 0
      let processedChunks = 0
    
      // Estimate chunks based on text length
      const estimatedChunks = Math.ceil(text.length / 100)
    
      return RunAnywhere.synthesizeStream(text, { rate: 1.0 }, (chunk) => {
        processedChunks++
        if (chunk.isFinal) {
          totalChunks = processedChunks
        }
        onProgress(Math.min(processedChunks / estimatedChunks, 1) * 100)
      })
    }
    
    // Usage
    const result = await synthesizeWithProgress(longText, (percent) =>
      console.log(`Progress: ${percent.toFixed(0)}%`)
    )
    

[​](https://docs.runanywhere.ai/react-native/tts/stream#performance-considerations)

Performance Considerations
-----------------------------------------------------------------------------------------------------------------

Streaming provides faster time-to-first-audio (TTFA) for long text. For short phrases (< 100 characters), the overhead may not be worth it — use `synthesize()` instead.

### 

[​](https://docs.runanywhere.ai/react-native/tts/stream#when-to-use-streaming)

When to Use Streaming

| Scenario | Recommended Method |
| --- | --- |
| Short phrases (< 50 chars) | `synthesize()` |
| Medium text (50-500 chars) | Either works |
| Long text (> 500 chars) | `synthesizeStream()` |
| Real-time AI responses | `synthesizeStream()` |
| Pre-generated content | `synthesize()` |

### 

[​](https://docs.runanywhere.ai/react-native/tts/stream#buffer-size)

Buffer Size

Chunks are typically 0.5-2 seconds of audio. This provides a good balance between:

*   **Latency**: Smaller chunks = faster first audio
*   **Efficiency**: Larger chunks = less overhead

[​](https://docs.runanywhere.ai/react-native/tts/stream#error-handling)

Error Handling
-----------------------------------------------------------------------------------------

Copy

Ask AI

    try {
      await RunAnywhere.synthesizeStream(text, { rate: 1.0 }, (chunk) => {
        try {
          playChunk(chunk)
        } catch (playError) {
          console.error('Playback error:', playError)
          // Continue streaming, don't fail the whole operation
        }
      })
    } catch (error) {
      if (isSDKError(error) && error.code === SDKErrorCode.ttsFailed) {
        console.error('TTS streaming failed:', error.message)
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/tts/stream#related)

Related
---------------------------------------------------------------------------

[Synthesize\
----------\
\
Basic synthesis](https://docs.runanywhere.ai/react-native/tts/synthesize)
[TTS Voices\
----------\
\
Available voices](https://docs.runanywhere.ai/react-native/tts/voices)
[Voice Agent\
-----------\
\
Full voice pipeline](https://docs.runanywhere.ai/react-native/voice-agent)

[synthesize()](https://docs.runanywhere.ai/react-native/tts/synthesize)
[Voices](https://docs.runanywhere.ai/react-native/tts/voices)

⌘I

---

# Configuration - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/configuration#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

React Native SDK

Configuration

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/configuration#overview)
    
*   [SDK Initialization](https://docs.runanywhere.ai/react-native/configuration#sdk-initialization)
    
*   [Basic Initialization](https://docs.runanywhere.ai/react-native/configuration#basic-initialization)
    
*   [Production Initialization](https://docs.runanywhere.ai/react-native/configuration#production-initialization)
    
*   [Full Configuration](https://docs.runanywhere.ai/react-native/configuration#full-configuration)
    
*   [Environment Modes](https://docs.runanywhere.ai/react-native/configuration#environment-modes)
    
*   [Logging](https://docs.runanywhere.ai/react-native/configuration#logging)
    
*   [Set Log Level](https://docs.runanywhere.ai/react-native/configuration#set-log-level)
    
*   [Log Levels](https://docs.runanywhere.ai/react-native/configuration#log-levels)
    
*   [Custom Logger (Advanced)](https://docs.runanywhere.ai/react-native/configuration#custom-logger-advanced)
    
*   [Events](https://docs.runanywhere.ai/react-native/configuration#events)
    
*   [EventBus](https://docs.runanywhere.ai/react-native/configuration#eventbus)
    
*   [Event Categories](https://docs.runanywhere.ai/react-native/configuration#event-categories)
    
*   [Event Subscription Examples](https://docs.runanywhere.ai/react-native/configuration#event-subscription-examples)
    
*   [SDK State](https://docs.runanywhere.ai/react-native/configuration#sdk-state)
    
*   [Check Initialization](https://docs.runanywhere.ai/react-native/configuration#check-initialization)
    
*   [Reset SDK](https://docs.runanywhere.ai/react-native/configuration#reset-sdk)
    
*   [Model Configuration](https://docs.runanywhere.ai/react-native/configuration#model-configuration)
    
*   [Generation Options](https://docs.runanywhere.ai/react-native/configuration#generation-options)
    
*   [STT Options](https://docs.runanywhere.ai/react-native/configuration#stt-options)
    
*   [TTS Options](https://docs.runanywhere.ai/react-native/configuration#tts-options)
    
*   [Storage Configuration](https://docs.runanywhere.ai/react-native/configuration#storage-configuration)
    
*   [Get Storage Info](https://docs.runanywhere.ai/react-native/configuration#get-storage-info)
    
*   [Clean Up Storage](https://docs.runanywhere.ai/react-native/configuration#clean-up-storage)
    
*   [React Native Specific](https://docs.runanywhere.ai/react-native/configuration#react-native-specific)
    
*   [App Lifecycle](https://docs.runanywhere.ai/react-native/configuration#app-lifecycle)
    
*   [Memory Warnings](https://docs.runanywhere.ai/react-native/configuration#memory-warnings)
    
*   [Model Management](https://docs.runanywhere.ai/react-native/configuration#model-management)
    
*   [Check Model Status](https://docs.runanywhere.ai/react-native/configuration#check-model-status)
    
*   [Audio Utilities](https://docs.runanywhere.ai/react-native/configuration#audio-utilities)
    
*   [Constants](https://docs.runanywhere.ai/react-native/configuration#constants)
    
*   [Recording](https://docs.runanywhere.ai/react-native/configuration#recording)
    
*   [Playback](https://docs.runanywhere.ai/react-native/configuration#playback)
    
*   [Audio Conversion](https://docs.runanywhere.ai/react-native/configuration#audio-conversion)
    
*   [Cleanup](https://docs.runanywhere.ai/react-native/configuration#cleanup)
    
*   [Helper](https://docs.runanywhere.ai/react-native/configuration#helper)
    
*   [Related](https://docs.runanywhere.ai/react-native/configuration#related)
    

[​](https://docs.runanywhere.ai/react-native/configuration#overview)

Overview
--------------------------------------------------------------------------------

This guide covers SDK initialization options, environment configuration, logging, and event subscriptions.

[​](https://docs.runanywhere.ai/react-native/configuration#sdk-initialization)

SDK Initialization
----------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/configuration#basic-initialization)

Basic Initialization

Copy

Ask AI

    import { RunAnywhere, SDKEnvironment } from '@runanywhere/core'
    
    // Development mode (no API key required)
    await RunAnywhere.initialize({
      environment: SDKEnvironment.Development,
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#production-initialization)

Production Initialization

Copy

Ask AI

    await RunAnywhere.initialize({
      apiKey: 'your-api-key',
      baseURL: 'https://api.runanywhere.ai',
      environment: SDKEnvironment.Production,
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#full-configuration)

Full Configuration

Copy

Ask AI

    interface SDKInitOptions {
      /** API key for authentication (production/staging) */
      apiKey?: string
    
      /** Base URL for API requests */
      baseURL?: string
    
      /** SDK environment */
      environment?: SDKEnvironment
    
      /** Supabase project URL (development mode only) */
      supabaseURL?: string
    
      /** Supabase anon key (development mode only) */
      supabaseKey?: string
    
      /** Enable debug logging */
      debug?: boolean
    }
    

[​](https://docs.runanywhere.ai/react-native/configuration#environment-modes)

Environment Modes
--------------------------------------------------------------------------------------------------

| Environment | Description | API Key | Logging |
| --- | --- | --- | --- |
| `Development` | Local development, full debugging | Optional | Debug |
| `Staging` | Testing with real services | Required | Info |
| `Production` | Production deployment | Required | Warning |

Copy

Ask AI

    enum SDKEnvironment {
      Development = 'development',
      Staging = 'staging',
      Production = 'production',
    }
    

[​](https://docs.runanywhere.ai/react-native/configuration#logging)

Logging
------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/configuration#set-log-level)

Set Log Level

Copy

Ask AI

    import { RunAnywhere, LogLevel } from '@runanywhere/core'
    
    // Set minimum log level
    RunAnywhere.setLogLevel(LogLevel.Debug)
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#log-levels)

Log Levels

| Level | Description | Use Case |
| --- | --- | --- |
| `Debug` | Detailed debugging info | Development |
| `Info` | General information | Staging |
| `Warning` | Potential issues | Production |
| `Error` | Errors and failures | Production |
| `Fault` | Critical failures | Always |

### 

[​](https://docs.runanywhere.ai/react-native/configuration#custom-logger-advanced)

Custom Logger (Advanced)

`SDKLogger` is used internally by the SDK. For most apps, the built-in logging controlled by `setLogLevel()` is sufficient.

Copy

Ask AI

    import { SDKLogger } from '@runanywhere/core'
    
    // Create a custom logger
    const logger = new SDKLogger('MyApp')
    
    logger.debug('Debug message', { modelId: 'llama-2' })
    logger.info('Info message')
    logger.warning('Warning message')
    logger.error('Error message', new Error('Something went wrong'))
    

[​](https://docs.runanywhere.ai/react-native/configuration#events)

Events
----------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/configuration#eventbus)

EventBus

The SDK provides a comprehensive event system for monitoring SDK activities.

Copy

Ask AI

    import { RunAnywhere, EventBus, EventCategory } from '@runanywhere/core'
    
    // Access EventBus via RunAnywhere
    RunAnywhere.events.onGeneration((event) => {
      console.log('Generation event:', event.type)
    })
    
    // Or use EventBus directly
    EventBus.on('Generation', (event) => {
      console.log('Event:', event)
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#event-categories)

Event Categories

| Category | Events |
| --- | --- |
| `Initialization` | `started`, `completed`, `failed` |
| `Generation` | `started`, `tokenGenerated`, `completed`, `failed`, `cancelled` |
| `Model` | `downloadStarted`, `downloadProgress`, `downloadCompleted`, `loadStarted`, `loadCompleted`, `unloaded` |
| `Voice` | `sttStarted`, `sttCompleted`, `ttsStarted`, `ttsCompleted`, `vadSpeechStarted`, `vadSpeechEnded` |
| `Storage` | `cleared`, `modelDeleted` |
| `Error` | Various error events |

### 

[​](https://docs.runanywhere.ai/react-native/configuration#event-subscription-examples)

Event Subscription Examples

Copy

Ask AI

    // Initialization events
    const unsubInit = RunAnywhere.events.onInitialization((event) => {
      switch (event.type) {
        case 'started':
          console.log('SDK initializing...')
          break
        case 'completed':
          console.log('SDK ready!')
          break
        case 'failed':
          console.error('Init failed:', event.error)
          break
      }
    })
    
    // Generation events
    const unsubGen = RunAnywhere.events.onGeneration((event) => {
      switch (event.type) {
        case 'started':
          console.log('Generation started')
          break
        case 'tokenGenerated':
          // Append token to your UI state
          setResponse((prev) => prev + event.token)
          break
        case 'completed':
          console.log('Done:', event.response.text)
          break
        case 'failed':
          console.error('Failed:', event.error)
          break
      }
    })
    
    // Model events
    const unsubModel = RunAnywhere.events.onModel((event) => {
      switch (event.type) {
        case 'downloadProgress':
          console.log(`${event.modelId}: ${(event.progress * 100).toFixed(1)}%`)
          break
        case 'downloadCompleted':
          console.log(`${event.modelId} downloaded`)
          break
        case 'loadCompleted':
          console.log(`${event.modelId} loaded`)
          break
      }
    })
    
    // Voice events
    const unsubVoice = RunAnywhere.events.onVoice((event) => {
      switch (event.type) {
        case 'sttStarted':
          console.log('Transcription started')
          break
        case 'sttCompleted':
          console.log('Transcript:', event.result.text)
          break
        case 'vadSpeechStarted':
          console.log('Speech detected')
          break
        case 'vadSpeechEnded':
          console.log('Speech ended')
          break
      }
    })
    
    // Clean up subscriptions
    unsubInit()
    unsubGen()
    unsubModel()
    unsubVoice()
    

[​](https://docs.runanywhere.ai/react-native/configuration#sdk-state)

SDK State
----------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/configuration#check-initialization)

Check Initialization

Copy

Ask AI

    // Check if SDK is initialized
    const isInit = await RunAnywhere.isInitialized()
    
    // Check properties
    console.log('Initialized:', RunAnywhere.isSDKInitialized)
    console.log('Services ready:', RunAnywhere.areServicesReady)
    console.log('Environment:', RunAnywhere.currentEnvironment)
    console.log('Version:', RunAnywhere.version)
    console.log('Device ID:', RunAnywhere.deviceId)
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#reset-sdk)

Reset SDK

Copy

Ask AI

    // Reset to uninitialized state
    await RunAnywhere.reset()
    
    // Or use destroy (alias)
    await RunAnywhere.destroy()
    
    // Can now reinitialize with different options
    await RunAnywhere.initialize({
      environment: SDKEnvironment.Production,
      apiKey: 'new-api-key',
    })
    

[​](https://docs.runanywhere.ai/react-native/configuration#model-configuration)

Model Configuration
------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/configuration#generation-options)

Generation Options

Copy

Ask AI

    const options: GenerationOptions = {
      maxTokens: 256, // Max tokens to generate
      temperature: 0.7, // Sampling temperature (0.0-2.0)
      topP: 0.95, // Nucleus sampling
      stopSequences: ['END'], // Stop sequences
      systemPrompt: 'You are helpful.', // System prompt
      preferredExecutionTarget: ExecutionTarget.OnDevice,
      preferredFramework: LLMFramework.LlamaCpp,
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#stt-options)

STT Options

Copy

Ask AI

    const sttOptions: STTOptions = {
      language: 'en', // Language code
      punctuation: true, // Add punctuation
      wordTimestamps: true, // Include word timing
      sampleRate: 16000, // Audio sample rate
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#tts-options)

TTS Options

Copy

Ask AI

    const ttsOptions: TTSConfiguration = {
      voice: 'en-US-female-1', // Voice ID
      rate: 1.0, // Speech rate (0.5-2.0)
      pitch: 1.0, // Pitch (0.5-2.0)
      volume: 1.0, // Volume (0.0-1.0)
    }
    

[​](https://docs.runanywhere.ai/react-native/configuration#storage-configuration)

Storage Configuration
----------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/configuration#get-storage-info)

Get Storage Info

Copy

Ask AI

    const storage = await RunAnywhere.getStorageInfo()
    
    console.log('Total space:', storage.totalSpace)
    console.log('Used by SDK:', storage.usedSpace)
    console.log('Free space:', storage.freeSpace)
    console.log('Models path:', storage.modelsPath)
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#clean-up-storage)

Clean Up Storage

Copy

Ask AI

    // Clear SDK cache
    await RunAnywhere.clearCache()
    
    // Clean temp files
    await RunAnywhere.cleanTempFiles()
    
    // Delete a specific model
    await RunAnywhere.deleteModel('model-id')
    

[​](https://docs.runanywhere.ai/react-native/configuration#react-native-specific)

React Native Specific
----------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/configuration#app-lifecycle)

App Lifecycle

Copy

Ask AI

    import { AppState, AppStateStatus } from 'react-native'
    
    // Handle app state changes
    useEffect(() => {
      const subscription = AppState.addEventListener('change', (state: AppStateStatus) => {
        if (state === 'background') {
          // Optionally unload models to free memory
          RunAnywhere.unloadModel()
        } else if (state === 'active') {
          // Reload model if needed
        }
      })
    
      return () => subscription.remove()
    }, [])
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#memory-warnings)

Memory Warnings

Use the SDK’s event system to handle low-memory situations:

Copy

Ask AI

    import { AppState } from 'react-native'
    import { RunAnywhere } from '@runanywhere/core'
    
    // Listen for OS memory warnings via the performance event bus
    RunAnywhere.events.onPerformance((event) => {
      if (event.type === 'memoryWarning') {
        console.log('Memory warning — usage:', event.usage)
        // Free memory by unloading unused models
        RunAnywhere.unloadModel()
      }
    })
    
    // Also listen for unexpected model unloads
    RunAnywhere.events.onModel((event) => {
      if (event.type === 'unloadCompleted') {
        console.log('Model was unloaded')
      }
    })
    
    // Proactively unload models when the app backgrounds
    useEffect(() => {
      const subscription = AppState.addEventListener('change', (state) => {
        if (state === 'background') {
          RunAnywhere.unloadModel()
        }
      })
      return () => subscription.remove()
    }, [])
    

[​](https://docs.runanywhere.ai/react-native/configuration#model-management)

Model Management
------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/configuration#check-model-status)

Check Model Status

Copy

Ask AI

    // Check if a specific model is loaded in memory
    const loaded = await RunAnywhere.isModelLoaded()
    
    // Check if a model has been downloaded
    const downloaded = await RunAnywhere.isModelDownloaded('smollm2-360m')
    
    // Get all registered models
    const models = await RunAnywhere.getAvailableModels()
    for (const model of models) {
      console.log(
        `${model.id}: ${model.name} (${model.isDownloaded ? 'downloaded' : 'not downloaded'})`
      )
    }
    
    // Get detailed info for a model
    const info = await RunAnywhere.getModelInfo('smollm2-360m')
    console.log('Local path:', info.localPath)
    

[​](https://docs.runanywhere.ai/react-native/configuration#audio-utilities)

Audio Utilities
----------------------------------------------------------------------------------------------

The SDK provides a built-in `RunAnywhere.Audio` utility for recording, playback, and audio conversion in React Native.

### 

[​](https://docs.runanywhere.ai/react-native/configuration#constants)

Constants

Copy

Ask AI

    RunAnywhere.Audio.SAMPLE_RATE // 16000 (STT default)
    RunAnywhere.Audio.TTS_SAMPLE_RATE // 22050 (TTS default)
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#recording)

Recording

Copy

Ask AI

    // Request microphone permission
    await RunAnywhere.Audio.requestPermission()
    
    // Start recording audio
    await RunAnywhere.Audio.startRecording()
    
    // Stop recording and get the recorded file
    const recording = await RunAnywhere.Audio.stopRecording()
    // recording.uri       - path to the recorded WAV file
    // recording.durationMs - recording duration in milliseconds
    
    // Cancel recording without saving
    await RunAnywhere.Audio.cancelRecording()
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#playback)

Playback

Copy

Ask AI

    // Play base64-encoded audio
    await RunAnywhere.Audio.playAudio(wavBase64)
    
    // Playback controls
    await RunAnywhere.Audio.stopPlayback()
    await RunAnywhere.Audio.pausePlayback()
    await RunAnywhere.Audio.resumePlayback()
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#audio-conversion)

Audio Conversion

Copy

Ask AI

    // Convert TTS output (float32 PCM) to playable WAV format
    const wavBase64 = RunAnywhere.Audio.createWavFromPCMFloat32(ttsResult.audio, ttsResult.sampleRate)
    
    // Then play it
    await RunAnywhere.Audio.playAudio(wavBase64)
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#cleanup)

Cleanup

Copy

Ask AI

    // Release audio resources
    await RunAnywhere.Audio.cleanup()
    

### 

[​](https://docs.runanywhere.ai/react-native/configuration#helper)

Helper

Copy

Ask AI

    // Format seconds to "mm:ss" string
    const formatted = RunAnywhere.Audio.formatDuration(125) // "2:05"
    

[​](https://docs.runanywhere.ai/react-native/configuration#related)

Related
------------------------------------------------------------------------------

[Error Handling\
--------------\
\
Handle errors gracefully](https://docs.runanywhere.ai/react-native/error-handling)
[Best Practices\
--------------\
\
Optimization tips](https://docs.runanywhere.ai/react-native/best-practices)

[Voice Agent](https://docs.runanywhere.ai/react-native/voice-agent)
[Error Handling](https://docs.runanywhere.ai/react-native/error-handling)

⌘I

---

# synthesize() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/tts/synthesize#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

synthesize()

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/tts/synthesize#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/react-native/tts/synthesize#basic-usage)
    
*   [Setup](https://docs.runanywhere.ai/react-native/tts/synthesize#setup)
    
*   [API Reference](https://docs.runanywhere.ai/react-native/tts/synthesize#api-reference)
    
*   [synthesize](https://docs.runanywhere.ai/react-native/tts/synthesize#synthesize)
    
*   [Configuration](https://docs.runanywhere.ai/react-native/tts/synthesize#configuration)
    
*   [Result](https://docs.runanywhere.ai/react-native/tts/synthesize#result)
    
*   [Examples](https://docs.runanywhere.ai/react-native/tts/synthesize#examples)
    
*   [Basic Synthesis](https://docs.runanywhere.ai/react-native/tts/synthesize#basic-synthesis)
    
*   [With Voice Options](https://docs.runanywhere.ai/react-native/tts/synthesize#with-voice-options)
    
*   [Play Audio](https://docs.runanywhere.ai/react-native/tts/synthesize#play-audio)
    
*   [Using System TTS](https://docs.runanywhere.ai/react-native/tts/synthesize#using-system-tts)
    
*   [Get Available Voices](https://docs.runanywhere.ai/react-native/tts/synthesize#get-available-voices)
    
*   [Converting Audio for Playback](https://docs.runanywhere.ai/react-native/tts/synthesize#converting-audio-for-playback)
    
*   [Voice Options Explained](https://docs.runanywhere.ai/react-native/tts/synthesize#voice-options-explained)
    
*   [Error Handling](https://docs.runanywhere.ai/react-native/tts/synthesize#error-handling)
    
*   [Related](https://docs.runanywhere.ai/react-native/tts/synthesize#related)
    

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#overview)

Overview
---------------------------------------------------------------------------------

The Text-to-Speech (TTS) API converts text to spoken audio using on-device neural voice synthesis with Piper TTS. All synthesis happens locally on the device.

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#basic-usage)

Basic Usage
---------------------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    // Synthesize speech
    const result = await RunAnywhere.synthesize('Hello, welcome to the RunAnywhere SDK.', {
      rate: 1.0,
      pitch: 1.0,
      volume: 1.0,
    })
    
    console.log('Duration:', result.duration, 'seconds')
    console.log('Sample rate:', result.sampleRate)
    // result.audio contains base64-encoded float32 PCM
    

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#setup)

Setup
---------------------------------------------------------------------------

Before synthesizing, download and load a TTS model:

Copy

Ask AI

    import { RunAnywhere, ModelCategory } from '@runanywhere/core'
    import { ONNX, ModelArtifactType } from '@runanywhere/onnx'
    
    // 1. Initialize SDK and register ONNX backend
    await RunAnywhere.initialize({ environment: SDKEnvironment.Development })
    ONNX.register()
    
    // 2. Add TTS model
    await ONNX.addModel({
      id: 'piper-en-lessac',
      name: 'Piper English (Lessac)',
      url: 'https://github.com/RunanywhereAI/sherpa-onnx/releases/download/runanywhere-models-v1/vits-piper-en_US-lessac-medium.tar.gz',
      modality: ModelCategory.SpeechSynthesis,
      artifactType: ModelArtifactType.TarGzArchive,
      memoryRequirement: 65_000_000,
    })
    
    // 3. Download model
    await RunAnywhere.downloadModel('piper-en-lessac', (progress) => {
      console.log(`Download: ${(progress.progress * 100).toFixed(1)}%`)
    })
    
    // 4. Load model
    const modelInfo = await RunAnywhere.getModelInfo('piper-en-lessac')
    await RunAnywhere.loadTTSModel(modelInfo.localPath, 'piper')
    

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#api-reference)

API Reference
-------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#synthesize)

`synthesize`

Convert text to audio data.

Copy

Ask AI

    await RunAnywhere.synthesize(
      text: string,
      options?: TTSConfiguration
    ): Promise<TTSResult>
    

**Parameters:**

| Parameter | Type | Description |
| --- | --- | --- |
| `text` | `string` | Text to synthesize |
| `options` | `TTSConfiguration` | Optional voice settings |

### 

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#configuration)

Configuration

Copy

Ask AI

    interface TTSConfiguration {
      /** Voice identifier */
      voice?: string
    
      /** Speech rate (0.5 - 2.0, default: 1.0) */
      rate?: number
    
      /** Pitch adjustment (0.5 - 2.0, default: 1.0) */
      pitch?: number
    
      /** Volume (0.0 - 1.0, default: 1.0) */
      volume?: number
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#result)

Result

Copy

Ask AI

    interface TTSResult {
      /** Base64-encoded audio (float32 PCM) */
      audio: string
    
      /** Audio sample rate in Hz */
      sampleRate: number
    
      /** Number of audio samples */
      numSamples: number
    
      /** Audio duration in seconds */
      duration: number
    }
    

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#examples)

Examples
---------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#basic-synthesis)

Basic Synthesis

Copy

Ask AI

    const result = await RunAnywhere.synthesize('Hello, world!')
    
    console.log('Audio duration:', result.duration, 'seconds')
    console.log('Sample rate:', result.sampleRate, 'Hz')
    console.log('Samples:', result.numSamples)
    

### 

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#with-voice-options)

With Voice Options

Copy

Ask AI

    // Slower and lower pitch
    const slow = await RunAnywhere.synthesize('This is spoken slowly with a lower pitch.', {
      rate: 0.75,
      pitch: 0.8,
      volume: 1.0,
    })
    
    // Faster and higher pitch
    const fast = await RunAnywhere.synthesize('This is spoken quickly with a higher pitch!', {
      rate: 1.5,
      pitch: 1.2,
      volume: 1.0,
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#play-audio)

Play Audio

TTSPlayer.tsx

Copy

Ask AI

    import React, { useState, useCallback } from 'react'
    import { View, Button, TextInput, Text } from 'react-native'
    import { RunAnywhere } from '@runanywhere/core'
    import Sound from 'react-native-sound' // Example audio playback library
    
    export function TTSPlayer() {
      const [text, setText] = useState('')
      const [isPlaying, setIsPlaying] = useState(false)
      const [duration, setDuration] = useState<number | null>(null)
    
      const handleSpeak = useCallback(async () => {
        if (!text.trim()) return
    
        setIsPlaying(true)
        try {
          const result = await RunAnywhere.synthesize(text, {
            rate: 1.0,
            pitch: 1.0,
          })
    
          setDuration(result.duration)
    
          // Convert base64 to audio and play
          const audioBuffer = base64ToArrayBuffer(result.audio)
          const sound = new Sound(audioBuffer, '', (error) => {
            if (error) {
              console.error('Failed to load sound', error)
              setIsPlaying(false)
              return
            }
            sound.play(() => {
              setIsPlaying(false)
              sound.release()
            })
          })
        } catch (error) {
          console.error('Synthesis failed:', error)
          setIsPlaying(false)
        }
      }, [text])
    
      return (
        <View style={{ padding: 16 }}>
          <TextInput
            value={text}
            onChangeText={setText}
            placeholder="Enter text to speak..."
            multiline
            style={{ borderWidth: 1, padding: 12, minHeight: 80 }}
          />
          <Button
            title={isPlaying ? 'Speaking...' : 'Speak'}
            onPress={handleSpeak}
            disabled={isPlaying || !text.trim()}
          />
          {duration && (
            <Text style={{ marginTop: 8, color: '#666' }}>Duration: {duration.toFixed(2)}s</Text>
          )}
        </View>
      )
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#using-system-tts)

Using System TTS

For simpler playback using platform’s built-in TTS:

Copy

Ask AI

    // Use system TTS (AVSpeechSynthesizer on iOS, Android TTS)
    await RunAnywhere.speak('Hello from system TTS!', {
      rate: 1.0,
      pitch: 1.0,
      volume: 1.0,
    })
    
    // Check if currently speaking
    const speaking = await RunAnywhere.isSpeaking()
    
    // Stop playback
    await RunAnywhere.stopSpeaking()
    

### 

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#get-available-voices)

Get Available Voices

Copy

Ask AI

    const voices = await RunAnywhere.availableTTSVoices()
    
    for (const voice of voices) {
      console.log(`${voice.id}: ${voice.name} (${voice.language})`)
    }
    
    // Use a specific voice
    await RunAnywhere.synthesize('Hello!', {
      voice: 'en-US-female-1',
    })
    

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#converting-audio-for-playback)

Converting Audio for Playback
---------------------------------------------------------------------------------------------------------------------------

The synthesized audio is base64-encoded float32 PCM. Use the SDK’s `Audio` utility to convert and play:

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    // Convert to WAV and play using SDK's Audio utility
    async function playTTSResult(result: TTSResult) {
      // Convert float32 PCM to playable WAV format
      const wavBase64 = RunAnywhere.Audio.createWavFromPCMFloat32(result.audio, result.sampleRate)
    
      // Play the audio
      await RunAnywhere.Audio.playAudio(wavBase64)
    }
    
    // Stop playback if needed
    await RunAnywhere.Audio.stopPlayback()
    

`AudioContext` / Web Audio API does **not** work in React Native. Always use `RunAnywhere.Audio.createWavFromPCMFloat32()` to convert TTS output before playback.

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#voice-options-explained)

Voice Options Explained
---------------------------------------------------------------------------------------------------------------

| Option | Range | Default | Effect |
| --- | --- | --- | --- |
| `rate` | 0.5 - 2.0 | 1.0 | Speech speed (1.0 = normal) |
| `pitch` | 0.5 - 2.0 | 1.0 | Voice pitch (1.0 = normal) |
| `volume` | 0.0 - 1.0 | 1.0 | Audio volume (1.0 = full) |

Copy

Ask AI

    // Different presets
    const presets = {
      normal: { rate: 1.0, pitch: 1.0, volume: 1.0 },
      slow: { rate: 0.7, pitch: 1.0, volume: 1.0 },
      fast: { rate: 1.4, pitch: 1.0, volume: 1.0 },
      deep: { rate: 1.0, pitch: 0.7, volume: 1.0 },
      high: { rate: 1.0, pitch: 1.3, volume: 1.0 },
      quiet: { rate: 1.0, pitch: 1.0, volume: 0.5 },
    }
    
    const result = await RunAnywhere.synthesize(text, presets.slow)
    

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#error-handling)

Error Handling
---------------------------------------------------------------------------------------------

Copy

Ask AI

    import { isSDKError, SDKErrorCode } from '@runanywhere/core'
    
    try {
      const result = await RunAnywhere.synthesize(text)
    } catch (error) {
      if (isSDKError(error)) {
        switch (error.code) {
          case SDKErrorCode.notInitialized:
            console.error('SDK not initialized')
            break
          case SDKErrorCode.modelNotLoaded:
            console.error('Load a TTS model first')
            break
          case SDKErrorCode.ttsFailed:
            console.error('Synthesis failed:', error.message)
            break
        }
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/tts/synthesize#related)

Related
-------------------------------------------------------------------------------

[TTS Streaming\
-------------\
\
Streaming synthesis](https://docs.runanywhere.ai/react-native/tts/stream)
[TTS Voices\
----------\
\
Available voices](https://docs.runanywhere.ai/react-native/tts/voices)
[Voice Agent\
-----------\
\
Full voice pipeline](https://docs.runanywhere.ai/react-native/voice-agent)

[STT Options](https://docs.runanywhere.ai/react-native/stt/options)
[Streaming TTS](https://docs.runanywhere.ai/react-native/tts/stream)

⌘I

---

# generateStream() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/llm/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

generateStream()

On this page

*   [Basic Usage](https://docs.runanywhere.ai/swift/llm/stream#basic-usage)
    
*   [Method Signature](https://docs.runanywhere.ai/swift/llm/stream#method-signature)
    
*   [Parameters](https://docs.runanywhere.ai/swift/llm/stream#parameters)
    
*   [Returns](https://docs.runanywhere.ai/swift/llm/stream#returns)
    
*   [SwiftUI Integration](https://docs.runanywhere.ai/swift/llm/stream#swiftui-integration)
    
*   [Basic Streaming View](https://docs.runanywhere.ai/swift/llm/stream#basic-streaming-view)
    
*   [With Typing Animation Effect](https://docs.runanywhere.ai/swift/llm/stream#with-typing-animation-effect)
    
*   [Cancellation](https://docs.runanywhere.ai/swift/llm/stream#cancellation)
    
*   [Error Handling](https://docs.runanywhere.ai/swift/llm/stream#error-handling)
    
*   [Check Streaming Support](https://docs.runanywhere.ai/swift/llm/stream#check-streaming-support)
    
*   [Performance Tips](https://docs.runanywhere.ai/swift/llm/stream#performance-tips)
    
*   [Complete Chat Example](https://docs.runanywhere.ai/swift/llm/stream#complete-chat-example)
    

The `generateStream()` method enables real-time token streaming, perfect for building responsive chat interfaces where text appears progressively.

[​](https://docs.runanywhere.ai/swift/llm/stream#basic-usage)

Basic Usage
----------------------------------------------------------------------------

Copy

Ask AI

    let result = try await RunAnywhere.generateStream(
        "Tell me a story about a brave knight",
        options: LLMGenerationOptions(maxTokens: 500)
    )
    
    // Display tokens as they arrive
    for try await token in result.stream {
        print(token, terminator: "")
    }
    
    // Get final metrics after streaming completes
    let metrics = try await result.result.value
    print("\n\nGenerated \(metrics.tokensUsed) tokens at \(metrics.tokensPerSecond) tok/s")
    

[​](https://docs.runanywhere.ai/swift/llm/stream#method-signature)

Method Signature
--------------------------------------------------------------------------------------

Copy

Ask AI

    public static func generateStream(
        _ prompt: String,
        options: LLMGenerationOptions? = nil
    ) async throws -> LLMStreamingResult
    

### 

[​](https://docs.runanywhere.ai/swift/llm/stream#parameters)

Parameters

| Parameter | Type | Description |
| --- | --- | --- |
| `prompt` | `String` | The text prompt |
| `options` | `LLMGenerationOptions?` | Generation configuration (optional) |

### 

[​](https://docs.runanywhere.ai/swift/llm/stream#returns)

Returns

An `LLMStreamingResult` containing:

Copy

Ask AI

    public struct LLMStreamingResult: Sendable {
        // Async stream of tokens
        public let stream: AsyncThrowingStream<String, Error>
    
        // Task that completes with final metrics
        public let result: Task<LLMGenerationResult, Error>
    }
    

[​](https://docs.runanywhere.ai/swift/llm/stream#swiftui-integration)

SwiftUI Integration
--------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/llm/stream#basic-streaming-view)

Basic Streaming View

Copy

Ask AI

    struct StreamingView: View {
        @State private var prompt = ""
        @State private var response = ""
        @State private var isStreaming = false
        @State private var tokensPerSecond: Double = 0
    
        var body: some View {
            VStack(spacing: 16) {
                // Prompt input
                TextField("Enter your prompt...", text: $prompt)
                    .textFieldStyle(.roundedBorder)
    
                // Generate button
                Button(action: { Task { await generate() } }) {
                    Label(
                        isStreaming ? "Generating..." : "Generate",
                        systemImage: isStreaming ? "stop.fill" : "play.fill"
                    )
                }
                .disabled(prompt.isEmpty)
    
                // Streaming response
                ScrollView {
                    Text(response)
                        .frame(maxWidth: .infinity, alignment: .leading)
                }
                .frame(maxHeight: 300)
                .background(Color.gray.opacity(0.1))
                .cornerRadius(8)
    
                // Metrics
                if tokensPerSecond > 0 {
                    Text("\(String(format: "%.1f", tokensPerSecond)) tokens/sec")
                        .font(.caption)
                        .foregroundColor(.secondary)
                }
            }
            .padding()
        }
    
        func generate() async {
            response = ""
            isStreaming = true
    
            do {
                let result = try await RunAnywhere.generateStream(
                    prompt,
                    options: LLMGenerationOptions(maxTokens: 500)
                )
    
                // Stream tokens to UI
                for try await token in result.stream {
                    await MainActor.run {
                        response += token
                    }
                }
    
                // Get final metrics
                let metrics = try await result.result.value
                await MainActor.run {
                    tokensPerSecond = metrics.tokensPerSecond
                    isStreaming = false
                }
    
            } catch {
                await MainActor.run {
                    response = "Error: \(error.localizedDescription)"
                    isStreaming = false
                }
            }
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/llm/stream#with-typing-animation-effect)

With Typing Animation Effect

Copy

Ask AI

    class StreamingViewModel: ObservableObject {
        @Published var displayedText = ""
        @Published var isStreaming = false
        @Published var metrics: LLMGenerationResult?
    
        private var fullText = ""
    
        func stream(prompt: String) async {
            await MainActor.run {
                displayedText = ""
                fullText = ""
                isStreaming = true
                metrics = nil
            }
    
            do {
                let result = try await RunAnywhere.generateStream(
                    prompt,
                    options: LLMGenerationOptions(maxTokens: 500, temperature: 0.7)
                )
    
                for try await token in result.stream {
                    fullText += token
    
                    // Update UI on main thread
                    await MainActor.run {
                        displayedText = fullText
                    }
                }
    
                let finalMetrics = try await result.result.value
                await MainActor.run {
                    metrics = finalMetrics
                    isStreaming = false
                }
    
            } catch {
                await MainActor.run {
                    displayedText = "Error: \(error.localizedDescription)"
                    isStreaming = false
                }
            }
        }
    
        func cancel() async {
            await RunAnywhere.cancelGeneration()
            await MainActor.run {
                isStreaming = false
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/llm/stream#cancellation)

Cancellation
------------------------------------------------------------------------------

Cancel streaming mid-generation:

Copy

Ask AI

    // Start streaming in a task
    let streamTask = Task {
        let result = try await RunAnywhere.generateStream(prompt)
        for try await token in result.stream {
            print(token, terminator: "")
        }
    }
    
    // Cancel after 5 seconds
    try await Task.sleep(for: .seconds(5))
    await RunAnywhere.cancelGeneration()
    streamTask.cancel()
    

[​](https://docs.runanywhere.ai/swift/llm/stream#error-handling)

Error Handling
----------------------------------------------------------------------------------

Copy

Ask AI

    do {
        let result = try await RunAnywhere.generateStream(prompt)
    
        for try await token in result.stream {
            // Handle each token
            await updateUI(with: token)
        }
    
        // Stream completed successfully
        let metrics = try await result.result.value
    
    } catch is CancellationError {
        print("Generation was cancelled")
    } catch let error as SDKError {
        switch error.code {
        case .generationFailed:
            print("Generation failed: \(error.message)")
        case .streamingNotSupported:
            print("Model doesn't support streaming")
        default:
            print("Error: \(error.localizedDescription)")
        }
    }
    

[​](https://docs.runanywhere.ai/swift/llm/stream#check-streaming-support)

Check Streaming Support
----------------------------------------------------------------------------------------------------

Not all models support streaming. Check before calling:

Copy

Ask AI

    if await RunAnywhere.supportsLLMStreaming {
        let result = try await RunAnywhere.generateStream(prompt)
        // ... handle stream
    } else {
        // Fall back to non-streaming
        let result = try await RunAnywhere.generate(prompt)
        print(result.text)
    }
    

[​](https://docs.runanywhere.ai/swift/llm/stream#performance-tips)

Performance Tips
--------------------------------------------------------------------------------------

Batch UI updates

Instead of updating the UI for every token, batch updates:

Copy

Ask AI

    var buffer = ""
    var lastUpdate = Date()
    
    for try await token in result.stream {
        buffer += token
        
        // Update UI every 50ms
        if Date().timeIntervalSince(lastUpdate) > 0.05 {
            await MainActor.run { displayedText += buffer }
            buffer = ""
            lastUpdate = Date()
        }
    }
    
    // Flush remaining buffer
    if !buffer.isEmpty {
        await MainActor.run { displayedText += buffer }
    }

Use attributedString for markdown

For markdown rendering, consider processing the complete text periodically rather than on every token.

Memory management

For very long generations, consider trimming displayed history to prevent memory issues.

[​](https://docs.runanywhere.ai/swift/llm/stream#complete-chat-example)

Complete Chat Example
------------------------------------------------------------------------------------------------

Copy

Ask AI

    struct ChatMessage: Identifiable {
        let id = UUID()
        let role: Role
        var content: String
    
        enum Role {
            case user, assistant
        }
    }
    
    @Observable
    class ChatViewModel {
        var messages: [ChatMessage] = []
        var isStreaming = false
    
        func send(_ text: String) async {
            // Add user message
            messages.append(ChatMessage(role: .user, content: text))
    
            // Add placeholder for assistant
            let assistantMessage = ChatMessage(role: .assistant, content: "")
            messages.append(assistantMessage)
            let messageIndex = messages.count - 1
    
            isStreaming = true
    
            do {
                let result = try await RunAnywhere.generateStream(
                    text,
                    options: LLMGenerationOptions(maxTokens: 500)
                )
    
                for try await token in result.stream {
                    await MainActor.run {
                        messages[messageIndex].content += token
                    }
                }
    
            } catch {
                await MainActor.run {
                    messages[messageIndex].content = "Error: \(error.localizedDescription)"
                }
            }
    
            isStreaming = false
        }
    }
    

[generate()\
----------\
\
Non-streaming generation →](https://docs.runanywhere.ai/swift/llm/generate)
[System Prompts\
--------------\
\
Control model behavior →](https://docs.runanywhere.ai/swift/llm/system-prompts)

[generate()](https://docs.runanywhere.ai/swift/llm/generate)
[System Prompts](https://docs.runanywhere.ai/swift/llm/system-prompts)

⌘I

---

# Streaming STT - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/stt/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

STT (Speech-to-Text)

Streaming STT

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/stt/stream#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/react-native/stt/stream#basic-usage)
    
*   [API Reference](https://docs.runanywhere.ai/react-native/stt/stream#api-reference)
    
*   [startSTTStream](https://docs.runanywhere.ai/react-native/stt/stream#startsttstream)
    
*   [Configuration](https://docs.runanywhere.ai/react-native/stt/stream#configuration)
    
*   [Session Methods](https://docs.runanywhere.ai/react-native/stt/stream#session-methods)
    
*   [Examples](https://docs.runanywhere.ai/react-native/stt/stream#examples)
    
*   [Real-Time Microphone Transcription](https://docs.runanywhere.ai/react-native/stt/stream#real-time-microphone-transcription)
    
*   [With VAD Integration](https://docs.runanywhere.ai/react-native/stt/stream#with-vad-integration)
    
*   [Custom Hook](https://docs.runanywhere.ai/react-native/stt/stream#custom-hook)
    
*   [Performance Considerations](https://docs.runanywhere.ai/react-native/stt/stream#performance-considerations)
    
*   [Optimal Chunk Size](https://docs.runanywhere.ai/react-native/stt/stream#optimal-chunk-size)
    
*   [Buffer Management](https://docs.runanywhere.ai/react-native/stt/stream#buffer-management)
    
*   [Error Handling](https://docs.runanywhere.ai/react-native/stt/stream#error-handling)
    
*   [Related](https://docs.runanywhere.ai/react-native/stt/stream#related)
    

[​](https://docs.runanywhere.ai/react-native/stt/stream#overview)

Overview
-----------------------------------------------------------------------------

Real-time STT streaming allows you to transcribe audio as it’s being recorded, providing immediate feedback to users. This is essential for voice interfaces where low latency is critical.

[​](https://docs.runanywhere.ai/react-native/stt/stream#basic-usage)

Basic Usage
-----------------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    // Start streaming transcription
    const session = await RunAnywhere.startSTTStream({
      language: 'en',
      onPartialResult: (partial) => {
        console.log('Partial:', partial.text)
      },
      onFinalResult: (final) => {
        console.log('Final:', final.text)
      },
    })
    
    // Feed audio chunks as they arrive
    session.feedAudio(audioChunk) // Float32 samples
    
    // When done
    const result = await session.stop()
    console.log('Complete transcription:', result.text)
    

[​](https://docs.runanywhere.ai/react-native/stt/stream#api-reference)

API Reference
---------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/stt/stream#startsttstream)

`startSTTStream`

Start a streaming transcription session.

Copy

Ask AI

    await RunAnywhere.startSTTStream(
      config: STTStreamConfig
    ): Promise<STTStreamSession>
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/stream#configuration)

Configuration

Copy

Ask AI

    interface STTStreamConfig {
      /** Language code (e.g., 'en', 'es') */
      language?: string
    
      /** Sample rate of incoming audio (default: 16000) */
      sampleRate?: number
    
      /** Callback for partial results */
      onPartialResult?: (result: STTPartialResult) => void
    
      /** Callback for final segment results */
      onFinalResult?: (result: STTResult) => void
    
      /** Callback for errors */
      onError?: (error: Error) => void
    
      /** Enable Voice Activity Detection */
      enableVAD?: boolean
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/stream#session-methods)

Session Methods

Copy

Ask AI

    interface STTStreamSession {
      /** Feed audio samples to the stream */
      feedAudio(samples: number[]): void
    
      /** Pause transcription */
      pause(): void
    
      /** Resume transcription */
      resume(): void
    
      /** Stop and get final result */
      stop(): Promise<STTResult>
    
      /** Check if session is active */
      isActive: boolean
    }
    

[​](https://docs.runanywhere.ai/react-native/stt/stream#examples)

Examples
-----------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/stt/stream#real-time-microphone-transcription)

Real-Time Microphone Transcription

LiveTranscription.tsx

Copy

Ask AI

    import React, { useState, useCallback, useRef, useEffect } from 'react'
    import { View, Button, Text } from 'react-native'
    import { RunAnywhere, STTStreamSession } from '@runanywhere/core'
    import AudioRecord from 'react-native-audio-record'
    
    export function LiveTranscription() {
      const [isListening, setIsListening] = useState(false)
      const [transcript, setTranscript] = useState('')
      const [partial, setPartial] = useState('')
      const sessionRef = useRef<STTStreamSession | null>(null)
    
      useEffect(() => {
        // Initialize audio recording
        AudioRecord.init({
          sampleRate: 16000,
          channels: 1,
          bitsPerSample: 16,
          audioSource: 6,
        })
    
        return () => {
          if (sessionRef.current?.isActive) {
            sessionRef.current.stop()
          }
        }
      }, [])
    
      const startListening = useCallback(async () => {
        setIsListening(true)
        setTranscript('')
        setPartial('')
    
        // Start STT stream
        sessionRef.current = await RunAnywhere.startSTTStream({
          language: 'en',
          enableVAD: true,
          onPartialResult: (result) => {
            setPartial(result.text)
          },
          onFinalResult: (result) => {
            setTranscript((prev) => prev + result.text + ' ')
            setPartial('')
          },
        })
    
        // Start audio recording and pipe to STT
        AudioRecord.start()
        AudioRecord.on('data', (data: string) => {
          // Convert base64 to float32 samples
          const samples = base64ToFloat32(data)
          sessionRef.current?.feedAudio(samples)
        })
      }, [])
    
      const stopListening = useCallback(async () => {
        setIsListening(false)
        AudioRecord.stop()
    
        if (sessionRef.current) {
          const finalResult = await sessionRef.current.stop()
          setTranscript((prev) => prev + finalResult.text)
          setPartial('')
        }
      }, [])
    
      return (
        <View style={{ padding: 16 }}>
          <Button
            title={isListening ? '🔴 Stop' : '🎤 Start'}
            onPress={isListening ? stopListening : startListening}
          />
          <View style={{ marginTop: 16 }}>
            <Text style={{ fontSize: 16 }}>
              {transcript}
              <Text style={{ color: '#888' }}>{partial}</Text>
            </Text>
          </View>
        </View>
      )
    }
    
    // Helper to convert base64 audio to float32 samples
    function base64ToFloat32(base64: string): number[] {
      const binary = atob(base64)
      const int16Array = new Int16Array(binary.length / 2)
      for (let i = 0; i < int16Array.length; i++) {
        int16Array[i] = binary.charCodeAt(i * 2) | (binary.charCodeAt(i * 2 + 1) << 8)
      }
      return Array.from(int16Array).map((x) => x / 32768.0)
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/stream#with-vad-integration)

With VAD Integration

When VAD is enabled, the stream automatically detects speech segments:

Copy

Ask AI

    const session = await RunAnywhere.startSTTStream({
      language: 'en',
      enableVAD: true,
      onPartialResult: (result) => {
        // Updates while user is speaking
        updateUI(result.text)
      },
      onFinalResult: (result) => {
        // Called when VAD detects end of speech
        console.log('User finished speaking:', result.text)
      },
    })
    
    // Audio is processed, VAD handles segmentation automatically
    session.feedAudio(audioSamples)
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/stream#custom-hook)

Custom Hook

useSTTStream.ts

Copy

Ask AI

    import { useState, useRef, useCallback } from 'react'
    import { RunAnywhere, STTStreamSession, STTResult } from '@runanywhere/core'
    
    export function useSTTStream() {
      const [isStreaming, setIsStreaming] = useState(false)
      const [partial, setPartial] = useState('')
      const [transcript, setTranscript] = useState('')
      const sessionRef = useRef<STTStreamSession | null>(null)
    
      const start = useCallback(async (language = 'en') => {
        setIsStreaming(true)
        setPartial('')
    
        sessionRef.current = await RunAnywhere.startSTTStream({
          language,
          enableVAD: true,
          onPartialResult: (result) => setPartial(result.text),
          onFinalResult: (result) => {
            setTranscript((prev) => prev + result.text + ' ')
            setPartial('')
          },
        })
      }, [])
    
      const feedAudio = useCallback((samples: number[]) => {
        sessionRef.current?.feedAudio(samples)
      }, [])
    
      const stop = useCallback(async (): Promise<STTResult | null> => {
        setIsStreaming(false)
        if (sessionRef.current) {
          const result = await sessionRef.current.stop()
          setTranscript((prev) => prev + result.text)
          setPartial('')
          return result
        }
        return null
      }, [])
    
      const reset = useCallback(() => {
        setTranscript('')
        setPartial('')
      }, [])
    
      return {
        isStreaming,
        partial,
        transcript,
        start,
        feedAudio,
        stop,
        reset,
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/stt/stream#performance-considerations)

Performance Considerations
-----------------------------------------------------------------------------------------------------------------

Feed audio in chunks of 100-500ms for optimal balance between latency and accuracy. Too small chunks increase overhead; too large chunks increase perceived latency.

### 

[​](https://docs.runanywhere.ai/react-native/stt/stream#optimal-chunk-size)

Optimal Chunk Size

Copy

Ask AI

    // Recommended chunk sizes for 16kHz audio
    const CHUNK_DURATION_MS = 200
    const SAMPLE_RATE = 16000
    const CHUNK_SIZE = (SAMPLE_RATE * CHUNK_DURATION_MS) / 1000 // 3200 samples
    

### 

[​](https://docs.runanywhere.ai/react-native/stt/stream#buffer-management)

Buffer Management

Copy

Ask AI

    // Buffer audio to send in optimal chunks
    class AudioBuffer {
      private buffer: number[] = []
      private chunkSize: number
      private onChunkReady: (chunk: number[]) => void
    
      constructor(chunkSize: number, onChunkReady: (chunk: number[]) => void) {
        this.chunkSize = chunkSize
        this.onChunkReady = onChunkReady
      }
    
      push(samples: number[]) {
        this.buffer.push(...samples)
    
        while (this.buffer.length >= this.chunkSize) {
          const chunk = this.buffer.splice(0, this.chunkSize)
          this.onChunkReady(chunk)
        }
      }
    
      flush() {
        if (this.buffer.length > 0) {
          this.onChunkReady(this.buffer)
          this.buffer = []
        }
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/stt/stream#error-handling)

Error Handling
-----------------------------------------------------------------------------------------

Copy

Ask AI

    const session = await RunAnywhere.startSTTStream({
      language: 'en',
      onError: (error) => {
        console.error('STT stream error:', error.message)
        // Handle gracefully - maybe restart the stream
      },
    })
    
    // Wrap feedAudio in try-catch for safety
    try {
      session.feedAudio(samples)
    } catch (error) {
      console.error('Failed to feed audio:', error)
    }
    

[​](https://docs.runanywhere.ai/react-native/stt/stream#related)

Related
---------------------------------------------------------------------------

[Transcribe\
----------\
\
Batch transcription](https://docs.runanywhere.ai/react-native/stt/transcribe)
[STT Options\
-----------\
\
Configuration options](https://docs.runanywhere.ai/react-native/stt/options)
[VAD\
---\
\
Voice Activity Detection](https://docs.runanywhere.ai/react-native/vad)
[Voice Agent\
-----------\
\
Full voice pipeline](https://docs.runanywhere.ai/react-native/voice-agent)

[transcribe()](https://docs.runanywhere.ai/react-native/stt/transcribe)
[STT Options](https://docs.runanywhere.ai/react-native/stt/options)

⌘I

---

# synthesize() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/tts/synthesize#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

synthesize()

On this page

*   [Basic Usage](https://docs.runanywhere.ai/swift/tts/synthesize#basic-usage)
    
*   [Setup](https://docs.runanywhere.ai/swift/tts/synthesize#setup)
    
*   [Method Signatures](https://docs.runanywhere.ai/swift/tts/synthesize#method-signatures)
    
*   [Basic Synthesis](https://docs.runanywhere.ai/swift/tts/synthesize#basic-synthesis)
    
*   [Speak (with automatic playback)](https://docs.runanywhere.ai/swift/tts/synthesize#speak-with-automatic-playback)
    
*   [TTSOutput](https://docs.runanywhere.ai/swift/tts/synthesize#ttsoutput)
    
*   [TTSOptions](https://docs.runanywhere.ai/swift/tts/synthesize#ttsoptions)
    
*   [Examples](https://docs.runanywhere.ai/swift/tts/synthesize#examples)
    
*   [Simple Text-to-Speech](https://docs.runanywhere.ai/swift/tts/synthesize#simple-text-to-speech)
    
*   [With Custom Options](https://docs.runanywhere.ai/swift/tts/synthesize#with-custom-options)
    
*   [Speak (Automatic Playback)](https://docs.runanywhere.ai/swift/tts/synthesize#speak-automatic-playback)
    
*   [Playing Audio Manually](https://docs.runanywhere.ai/swift/tts/synthesize#playing-audio-manually)
    
*   [SwiftUI Voice Output](https://docs.runanywhere.ai/swift/tts/synthesize#swiftui-voice-output)
    
*   [Save to File](https://docs.runanywhere.ai/swift/tts/synthesize#save-to-file)
    
*   [Voice Management](https://docs.runanywhere.ai/swift/tts/synthesize#voice-management)
    
*   [Synthesis Metadata](https://docs.runanywhere.ai/swift/tts/synthesize#synthesis-metadata)
    
*   [Error Handling](https://docs.runanywhere.ai/swift/tts/synthesize#error-handling)
    
*   [Best Practices](https://docs.runanywhere.ai/swift/tts/synthesize#best-practices)
    

The `synthesize()` method converts text to natural-sounding speech using on-device neural voice models.

[​](https://docs.runanywhere.ai/swift/tts/synthesize#basic-usage)

Basic Usage
--------------------------------------------------------------------------------

Copy

Ask AI

    // Synthesize text to audio
    let output = try await RunAnywhere.synthesize("Hello! Welcome to RunAnywhere.")
    
    // Play the audio
    let player = try AVAudioPlayer(data: output.audioData)
    player.play()
    

[​](https://docs.runanywhere.ai/swift/tts/synthesize#setup)

Setup
--------------------------------------------------------------------

Before synthesizing, register the ONNX module and load a TTS voice:

Copy

Ask AI

    import RunAnywhere
    import ONNXRuntime
    
    // Register ONNX module (at app launch)
    @MainActor
    func setup() {
        ONNX.register()
    }
    
    // Load TTS voice
    try await RunAnywhere.loadTTSVoice("piper-en-us-amy")
    

[​](https://docs.runanywhere.ai/swift/tts/synthesize#method-signatures)

Method Signatures
--------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/tts/synthesize#basic-synthesis)

Basic Synthesis

Copy

Ask AI

    public static func synthesize(
        _ text: String,
        options: TTSOptions = TTSOptions()
    ) async throws -> TTSOutput
    

### 

[​](https://docs.runanywhere.ai/swift/tts/synthesize#speak-with-automatic-playback)

Speak (with automatic playback)

Copy

Ask AI

    public static func speak(
        _ text: String,
        options: TTSOptions = TTSOptions()
    ) async throws -> TTSSpeakResult
    

[​](https://docs.runanywhere.ai/swift/tts/synthesize#ttsoutput)

TTSOutput
----------------------------------------------------------------------------

Copy

Ask AI

    public struct TTSOutput: Sendable {
        public let audioData: Data                    // Audio bytes
        public let format: AudioFormat                // WAV, PCM, etc.
        public let duration: TimeInterval             // Audio duration
        public let phonemeTimestamps: [PhonemeTimestamp]? // Phoneme timing
        public let metadata: TTSSynthesisMetadata?    // Processing info
    }
    

[​](https://docs.runanywhere.ai/swift/tts/synthesize#ttsoptions)

TTSOptions
------------------------------------------------------------------------------

Copy

Ask AI

    public struct TTSOptions: Sendable {
        public let rate: Float          // Speed (0.5-2.0, default: 1.0)
        public let pitch: Float         // Pitch (0.5-2.0, default: 1.0)
        public let volume: Float        // Volume (0.0-1.0, default: 1.0)
        public let language: String     // Language code
        public let sampleRate: Int      // Output sample rate
        public let audioFormat: AudioFormat  // Output format
    
        public init(
            rate: Float = 1.0,
            pitch: Float = 1.0,
            volume: Float = 1.0,
            language: String = "en-US",
            sampleRate: Int = 22050,
            audioFormat: AudioFormat = .wav
        )
    }
    

[​](https://docs.runanywhere.ai/swift/tts/synthesize#examples)

Examples
--------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/tts/synthesize#simple-text-to-speech)

Simple Text-to-Speech

Copy

Ask AI

    let output = try await RunAnywhere.synthesize("Hello world!")
    print("Generated \(output.duration) seconds of audio")
    

### 

[​](https://docs.runanywhere.ai/swift/tts/synthesize#with-custom-options)

With Custom Options

Copy

Ask AI

    let options = TTSOptions(
        rate: 0.9,      // Slightly slower
        pitch: 1.1,     // Slightly higher pitch
        volume: 0.8     // 80% volume
    )
    
    let output = try await RunAnywhere.synthesize(
        "This is a customized voice.",
        options: options
    )
    

### 

[​](https://docs.runanywhere.ai/swift/tts/synthesize#speak-automatic-playback)

Speak (Automatic Playback)

Copy

Ask AI

    // Synthesize and play automatically
    try await RunAnywhere.speak("Hello! How can I help you today?")
    
    // Check if still speaking
    let isSpeaking = await RunAnywhere.isSpeaking
    
    // Stop playback
    await RunAnywhere.stopSpeaking()
    

### 

[​](https://docs.runanywhere.ai/swift/tts/synthesize#playing-audio-manually)

Playing Audio Manually

Copy

Ask AI

    import AVFoundation
    
    class SpeechPlayer {
        private var audioPlayer: AVAudioPlayer?
    
        func speak(_ text: String) async throws {
            let output = try await RunAnywhere.synthesize(text)
    
            audioPlayer = try AVAudioPlayer(data: output.audioData)
            audioPlayer?.play()
        }
    
        func stop() {
            audioPlayer?.stop()
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/tts/synthesize#swiftui-voice-output)

SwiftUI Voice Output

Copy

Ask AI

    struct TextToSpeechView: View {
        @State private var text = ""
        @State private var isSpeaking = false
        @State private var rate: Float = 1.0
    
        var body: some View {
            VStack(spacing: 20) {
                // Text input
                TextEditor(text: $text)
                    .frame(height: 150)
                    .border(Color.gray.opacity(0.3))
    
                // Rate slider
                VStack(alignment: .leading) {
                    Text("Speed: \(String(format: "%.1f", rate))x")
                    Slider(value: $rate, in: 0.5...2.0)
                }
    
                // Control buttons
                HStack(spacing: 20) {
                    Button(action: speak) {
                        Label("Speak", systemImage: "play.fill")
                            .padding()
                            .background(Color.blue)
                            .foregroundColor(.white)
                            .cornerRadius(10)
                    }
                    .disabled(text.isEmpty || isSpeaking)
    
                    Button(action: stop) {
                        Label("Stop", systemImage: "stop.fill")
                            .padding()
                            .background(Color.red)
                            .foregroundColor(.white)
                            .cornerRadius(10)
                    }
                    .disabled(!isSpeaking)
                }
            }
            .padding()
        }
    
        func speak() {
            isSpeaking = true
    
            Task {
                do {
                    try await RunAnywhere.speak(
                        text,
                        options: TTSOptions(rate: rate)
                    )
                    await MainActor.run {
                        isSpeaking = false
                    }
                } catch {
                    print("Speech failed: \(error)")
                    await MainActor.run {
                        isSpeaking = false
                    }
                }
            }
        }
    
        func stop() {
            Task {
                await RunAnywhere.stopSpeaking()
                await MainActor.run {
                    isSpeaking = false
                }
            }
        }
    }
    

### 

[​](https://docs.runanywhere.ai/swift/tts/synthesize#save-to-file)

Save to File

Copy

Ask AI

    func saveToFile(text: String, filename: String) async throws -> URL {
        let output = try await RunAnywhere.synthesize(text)
    
        let documentsPath = FileManager.default.urls(
            for: .documentDirectory,
            in: .userDomainMask
        ).first!
    
        let fileURL = documentsPath.appendingPathComponent("\(filename).wav")
        try output.audioData.write(to: fileURL)
    
        return fileURL
    }
    
    // Usage
    let url = try await saveToFile(text: "Hello world", filename: "greeting")
    print("Saved to: \(url.path)")
    

[​](https://docs.runanywhere.ai/swift/tts/synthesize#voice-management)

Voice Management
------------------------------------------------------------------------------------------

Copy

Ask AI

    // Load a voice
    try await RunAnywhere.loadTTSVoice("piper-en-us-amy")
    
    // Check if voice is loaded
    let isLoaded = await RunAnywhere.isTTSVoiceLoaded
    
    // Get current voice
    let voiceId = await RunAnywhere.currentTTSVoiceId
    
    // List available voices
    let voices = await RunAnywhere.availableTTSVoices
    for voice in voices {
        print(voice)
    }
    
    // Unload voice
    try await RunAnywhere.unloadTTSVoice()
    

[​](https://docs.runanywhere.ai/swift/tts/synthesize#synthesis-metadata)

Synthesis Metadata
----------------------------------------------------------------------------------------------

Copy

Ask AI

    let output = try await RunAnywhere.synthesize(text)
    
    if let metadata = output.metadata {
        print("Voice: \(metadata.voice)")
        print("Processing time: \(metadata.processingTime)s")
        print("Characters: \(metadata.characterCount)")
    
        let charsPerSecond = Double(metadata.characterCount) / metadata.processingTime
        print("Speed: \(String(format: "%.0f", charsPerSecond)) chars/sec")
    }
    

[​](https://docs.runanywhere.ai/swift/tts/synthesize#error-handling)

Error Handling
--------------------------------------------------------------------------------------

Copy

Ask AI

    do {
        let output = try await RunAnywhere.synthesize(text)
        playAudio(output.audioData)
    } catch let error as SDKError {
        switch error.code {
        case .modelNotFound:
            print("Load a TTS voice first")
        case .processingFailed:
            print("Synthesis failed: \(error.message)")
        case .invalidInput:
            print("Invalid text input")
        default:
            print("Error: \(error.localizedDescription)")
        }
    }
    

[​](https://docs.runanywhere.ai/swift/tts/synthesize#best-practices)

Best Practices
--------------------------------------------------------------------------------------

Chunk long text

For long text, use streaming synthesis or split into sentences to reduce latency.

Preload voices

Load TTS voices at app startup to avoid delays when synthesizing.

Handle audio session

Configure `AVAudioSession` appropriately for your app’s audio needs.

[Streaming TTS\
-------------\
\
Stream audio for long text →](https://docs.runanywhere.ai/swift/tts/stream)
[Voices\
------\
\
Available voices →](https://docs.runanywhere.ai/swift/tts/voices)

[STT Options](https://docs.runanywhere.ai/swift/stt/options)
[Streaming TTS](https://docs.runanywhere.ai/swift/tts/stream)

⌘I

---

# generateStream() - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/llm/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

LLM (Language Models)

generateStream()

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/llm/stream#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/react-native/llm/stream#basic-usage)
    
*   [API Reference](https://docs.runanywhere.ai/react-native/llm/stream#api-reference)
    
*   [Parameters](https://docs.runanywhere.ai/react-native/llm/stream#parameters)
    
*   [Returns](https://docs.runanywhere.ai/react-native/llm/stream#returns)
    
*   [Examples](https://docs.runanywhere.ai/react-native/llm/stream#examples)
    
*   [React Native Component](https://docs.runanywhere.ai/react-native/llm/stream#react-native-component)
    
*   [Custom Streaming Hook](https://docs.runanywhere.ai/react-native/llm/stream#custom-streaming-hook)
    
*   [With Typing Animation Effect](https://docs.runanywhere.ai/react-native/llm/stream#with-typing-animation-effect)
    
*   [Event-Based Streaming](https://docs.runanywhere.ai/react-native/llm/stream#event-based-streaming)
    
*   [Performance Tips](https://docs.runanywhere.ai/react-native/llm/stream#performance-tips)
    
*   [Optimize UI Updates](https://docs.runanywhere.ai/react-native/llm/stream#optimize-ui-updates)
    
*   [Related](https://docs.runanywhere.ai/react-native/llm/stream#related)
    

[​](https://docs.runanywhere.ai/react-native/llm/stream#overview)

Overview
-----------------------------------------------------------------------------

Token streaming allows you to display AI responses as they’re generated, token by token. This provides a much better user experience than waiting for the entire response, especially for longer outputs.

[​](https://docs.runanywhere.ai/react-native/llm/stream#basic-usage)

Basic Usage
-----------------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    const streamResult = await RunAnywhere.generateStream('Write a short story about a robot', {
      maxTokens: 200,
    })
    
    // Accumulate and display tokens as they arrive
    let fullResponse = ''
    for await (const token of streamResult.stream) {
      fullResponse += token
      setResponse(fullResponse) // Update UI state
    }
    
    // Get final metrics
    const finalResult = await streamResult.result
    console.log('Speed:', finalResult.tokensPerSecond, 'tok/s')
    

[​](https://docs.runanywhere.ai/react-native/llm/stream#api-reference)

API Reference
---------------------------------------------------------------------------------------

Copy

Ask AI

    await RunAnywhere.generateStream(
      prompt: string,
      options?: GenerationOptions
    ): Promise<LLMStreamingResult>
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/stream#parameters)

Parameters

Same as `generate()` - see [Generation Options](https://docs.runanywhere.ai/react-native/llm/generate#generation-options)
.

### 

[​](https://docs.runanywhere.ai/react-native/llm/stream#returns)

Returns

Copy

Ask AI

    interface LLMStreamingResult {
      /** Async iterator yielding tokens one at a time */
      stream: AsyncIterable<string>
    
      /** Promise resolving to final result with metrics */
      result: Promise<LLMGenerationResult>
    
      /** Cancel the ongoing generation */
      cancel: () => void
    }
    
    interface LLMGenerationResult {
      text: string
      tokensUsed: number
      modelUsed: string
      latencyMs: number
      framework: string
      tokensPerSecond: number
      timeToFirstTokenMs?: number
      responseTokens: number
      thinkingContent?: string
      thinkingTokens: number
      inputTokens: number
    }
    

`LLMGenerationResult` (returned by streaming) differs from `GenerationResult` (returned by `generate()`). Streaming results have `tokensPerSecond` at the top level, while `GenerationResult` nests it under `performanceMetrics`.

[​](https://docs.runanywhere.ai/react-native/llm/stream#examples)

Examples
-----------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/llm/stream#react-native-component)

React Native Component

StreamingChat.tsx

Copy

Ask AI

    import React, { useState, useCallback } from 'react'
    import { View, Text, Button, ScrollView } from 'react-native'
    import { RunAnywhere } from '@runanywhere/core'
    
    export function StreamingChat() {
      const [response, setResponse] = useState('')
      const [isStreaming, setIsStreaming] = useState(false)
      const [metrics, setMetrics] = useState<string>('')
    
      const handleStream = useCallback(async () => {
        setResponse('')
        setMetrics('')
        setIsStreaming(true)
    
        try {
          const streamResult = await RunAnywhere.generateStream('Explain how neural networks work', {
            maxTokens: 300,
            temperature: 0.7,
          })
    
          // Accumulate tokens
          let fullText = ''
          for await (const token of streamResult.stream) {
            fullText += token
            setResponse(fullText)
          }
    
          // Show final metrics
          const final = await streamResult.result
          setMetrics(
            `${final.tokensPerSecond?.toFixed(1)} tok/s | ` +
              `${final.latencyMs}ms | ${final.tokensUsed} tokens`
          )
        } catch (error) {
          setResponse('Error: ' + (error as Error).message)
        } finally {
          setIsStreaming(false)
        }
      }, [])
    
      return (
        <View style={{ flex: 1, padding: 16 }}>
          <Button
            title={isStreaming ? 'Streaming...' : 'Start Streaming'}
            onPress={handleStream}
            disabled={isStreaming}
          />
    
          <ScrollView style={{ flex: 1, marginTop: 16 }}>
            <Text style={{ fontSize: 16, lineHeight: 24 }}>
              {response}
              {isStreaming && <Text style={{ opacity: 0.5 }}>▊</Text>}
            </Text>
          </ScrollView>
    
          {metrics && <Text style={{ marginTop: 8, color: '#666', fontSize: 12 }}>{metrics}</Text>}
        </View>
      )
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/stream#custom-streaming-hook)

Custom Streaming Hook

useStreamingGenerate.ts

Copy

Ask AI

    import { useState, useCallback, useRef } from 'react'
    import { RunAnywhere, GenerationOptions, LLMGenerationResult } from '@runanywhere/core'
    
    interface StreamingState {
      text: string
      isStreaming: boolean
      error: Error | null
      metrics: LLMGenerationResult | null
    }
    
    export function useStreamingGenerate() {
      const [state, setState] = useState<StreamingState>({
        text: '',
        isStreaming: false,
        error: null,
        metrics: null,
      })
    
      const abortRef = useRef(false)
    
      const stream = useCallback(async (prompt: string, options?: GenerationOptions) => {
        abortRef.current = false
        setState({ text: '', isStreaming: true, error: null, metrics: null })
    
        try {
          const result = await RunAnywhere.generateStream(prompt, options)
    
          let accumulated = ''
          for await (const token of result.stream) {
            if (abortRef.current) break
            accumulated += token
            setState((s) => ({ ...s, text: accumulated }))
          }
    
          const finalMetrics = await result.result
          setState((s) => ({ ...s, isStreaming: false, metrics: finalMetrics }))
    
          return finalMetrics
        } catch (error) {
          const e = error instanceof Error ? error : new Error('Streaming failed')
          setState((s) => ({ ...s, isStreaming: false, error: e }))
          throw e
        }
      }, [])
    
      const cancel = useCallback(async () => {
        abortRef.current = true
        await RunAnywhere.cancelGeneration()
        setState((s) => ({ ...s, isStreaming: false }))
      }, [])
    
      return { ...state, stream, cancel }
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/llm/stream#with-typing-animation-effect)

With Typing Animation Effect

TypingEffect.tsx

Copy

Ask AI

    import React, { useState, useEffect, useCallback } from 'react'
    import { Text, View, Button } from 'react-native'
    import { RunAnywhere } from '@runanywhere/core'
    
    export function TypingEffect() {
      const [tokens, setTokens] = useState<string[]>([])
      const [displayedText, setDisplayedText] = useState('')
      const [isStreaming, setIsStreaming] = useState(false)
    
      // Animate tokens with a slight delay for effect
      useEffect(() => {
        if (tokens.length === 0) return
    
        let index = 0
        const interval = setInterval(() => {
          if (index < tokens.length) {
            setDisplayedText(tokens.slice(0, index + 1).join(''))
            index++
          } else {
            clearInterval(interval)
          }
        }, 20) // 20ms per token
    
        return () => clearInterval(interval)
      }, [tokens])
    
      const handleGenerate = useCallback(async () => {
        setTokens([])
        setDisplayedText('')
        setIsStreaming(true)
    
        const result = await RunAnywhere.generateStream('What makes a good programmer?', {
          maxTokens: 150,
        })
    
        const allTokens: string[] = []
        for await (const token of result.stream) {
          allTokens.push(token)
        }
    
        setTokens(allTokens)
        setIsStreaming(false)
      }, [])
    
      return (
        <View style={{ padding: 16 }}>
          <Button
            title={isStreaming ? 'Generating...' : 'Generate'}
            onPress={handleGenerate}
            disabled={isStreaming}
          />
          <Text style={{ marginTop: 16, fontSize: 16 }}>
            {displayedText}
            {isStreaming && '▊'}
          </Text>
        </View>
      )
    }
    

[​](https://docs.runanywhere.ai/react-native/llm/stream#event-based-streaming)

Event-Based Streaming
-------------------------------------------------------------------------------------------------------

You can also use the EventBus for streaming:

Copy

Ask AI

    import { RunAnywhere, EventBus } from '@runanywhere/core'
    
    // Subscribe to generation events
    const unsubscribe = RunAnywhere.events.onGeneration((event) => {
      switch (event.type) {
        case 'started':
          console.log('Generation started')
          break
        case 'tokenGenerated':
          // Append token to your UI state
          setResponse((prev) => prev + event.token)
          break
        case 'completed':
          console.log('Done:', event.response.tokensUsed, 'tokens')
          break
        case 'failed':
          console.error('Error:', event.error)
          break
        case 'cancelled':
          console.log('Generation cancelled')
          break
      }
    })
    
    // Generate (tokens will be emitted via events)
    await RunAnywhere.generate('Write a poem', { maxTokens: 100 })
    
    // Clean up
    unsubscribe()
    

[​](https://docs.runanywhere.ai/react-native/llm/stream#performance-tips)

Performance Tips
---------------------------------------------------------------------------------------------

Streaming has minimal overhead compared to non-streaming generation. The time-to-first-token (TTFT) is the same, and total generation time is nearly identical.

### 

[​](https://docs.runanywhere.ai/react-native/llm/stream#optimize-ui-updates)

Optimize UI Updates

For very fast generation, batch UI updates to avoid overwhelming React:

Copy

Ask AI

    import { useCallback, useRef, useState } from 'react'
    
    function useThrottledText(delay = 16) {
      // ~60fps
      const [text, setText] = useState('')
      const pendingRef = useRef('')
      const frameRef = useRef<number>()
    
      const appendToken = useCallback((token: string) => {
        pendingRef.current += token
    
        if (!frameRef.current) {
          frameRef.current = requestAnimationFrame(() => {
            setText(pendingRef.current)
            frameRef.current = undefined
          })
        }
      }, [])
    
      const reset = useCallback(() => {
        pendingRef.current = ''
        setText('')
        if (frameRef.current) {
          cancelAnimationFrame(frameRef.current)
          frameRef.current = undefined
        }
      }, [])
    
      return { text, appendToken, reset }
    }
    

[​](https://docs.runanywhere.ai/react-native/llm/stream#related)

Related
---------------------------------------------------------------------------

[Chat\
----\
\
Simple one-liner interface](https://docs.runanywhere.ai/react-native/llm/chat)
[Generate\
--------\
\
Full generation with metrics](https://docs.runanywhere.ai/react-native/llm/generate)
[System Prompts\
--------------\
\
Control AI behavior](https://docs.runanywhere.ai/react-native/llm/system-prompts)
[Events\
------\
\
Event subscription system](https://docs.runanywhere.ai/react-native/configuration#events)

[generate()](https://docs.runanywhere.ai/react-native/llm/generate)
[System Prompts](https://docs.runanywhere.ai/react-native/llm/system-prompts)

⌘I

---

# Voice Activity Detection - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/vad#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

React Native SDK

Voice Activity Detection

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/vad#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/react-native/vad#basic-usage)
    
*   [Setup](https://docs.runanywhere.ai/react-native/vad#setup)
    
*   [API Reference](https://docs.runanywhere.ai/react-native/vad#api-reference)
    
*   [initializeVAD](https://docs.runanywhere.ai/react-native/vad#initializevad)
    
*   [Configuration](https://docs.runanywhere.ai/react-native/vad#configuration)
    
*   [processVAD](https://docs.runanywhere.ai/react-native/vad#processvad)
    
*   [VADResult](https://docs.runanywhere.ai/react-native/vad#vadresult)
    
*   [Continuous VAD](https://docs.runanywhere.ai/react-native/vad#continuous-vad)
    
*   [SpeechActivityEvent](https://docs.runanywhere.ai/react-native/vad#speechactivityevent)
    
*   [Examples](https://docs.runanywhere.ai/react-native/vad#examples)
    
*   [Single Frame Processing](https://docs.runanywhere.ai/react-native/vad#single-frame-processing)
    
*   [Continuous Listening](https://docs.runanywhere.ai/react-native/vad#continuous-listening)
    
*   [VAD with Recording](https://docs.runanywhere.ai/react-native/vad#vad-with-recording)
    
*   [Push-to-Talk Alternative](https://docs.runanywhere.ai/react-native/vad#push-to-talk-alternative)
    
*   [Configuration Tips](https://docs.runanywhere.ai/react-native/vad#configuration-tips)
    
*   [Energy Threshold](https://docs.runanywhere.ai/react-native/vad#energy-threshold)
    
*   [Auto Calibration](https://docs.runanywhere.ai/react-native/vad#auto-calibration)
    
*   [Error Handling](https://docs.runanywhere.ai/react-native/vad#error-handling)
    
*   [Related](https://docs.runanywhere.ai/react-native/vad#related)
    

[​](https://docs.runanywhere.ai/react-native/vad#overview)

Overview
----------------------------------------------------------------------

Voice Activity Detection (VAD) determines when speech is present in an audio stream. It’s essential for building voice interfaces that can automatically detect when users start and stop speaking.

[​](https://docs.runanywhere.ai/react-native/vad#basic-usage)

Basic Usage
----------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    // Initialize VAD
    await RunAnywhere.initializeVAD({
      energyThreshold: 0.5,
      sampleRate: 16000,
    })
    
    // Process audio samples
    const result = await RunAnywhere.processVAD(audioSamples)
    
    if (result.isSpeech) {
      console.log('Speech detected!', result.confidence)
    }
    

[​](https://docs.runanywhere.ai/react-native/vad#setup)

Setup
----------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere, ModelCategory } from '@runanywhere/core'
    import { ONNX, ModelArtifactType } from '@runanywhere/onnx'
    
    // 1. Initialize SDK and register ONNX backend
    await RunAnywhere.initialize({ environment: SDKEnvironment.Development })
    ONNX.register()
    
    // 2. Add Silero VAD model
    await ONNX.addModel({
      id: 'silero-vad',
      name: 'Silero VAD',
      url: 'https://github.com/RunanywhereAI/sherpa-onnx/releases/.../silero-vad.tar.gz',
      modality: ModelCategory.Audio,
      artifactType: ModelArtifactType.TarGzArchive,
      memoryRequirement: 5_000_000,
    })
    
    // 3. Download and load
    await RunAnywhere.downloadModel('silero-vad')
    const modelInfo = await RunAnywhere.getModelInfo('silero-vad')
    await RunAnywhere.loadVADModel(modelInfo.localPath)
    

[​](https://docs.runanywhere.ai/react-native/vad#api-reference)

API Reference
--------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/vad#initializevad)

`initializeVAD`

Configure VAD settings.

Copy

Ask AI

    await RunAnywhere.initializeVAD(config?: VADConfiguration): Promise<boolean>
    

### 

[​](https://docs.runanywhere.ai/react-native/vad#configuration)

Configuration

Copy

Ask AI

    interface VADConfiguration {
      /** Energy threshold for speech detection (0.0-1.0) */
      energyThreshold?: number
    
      /** Audio sample rate in Hz (default: 16000) */
      sampleRate?: number
    
      /** Frame length in milliseconds (default: 30) */
      frameLength?: number
    
      /** Enable auto-calibration */
      autoCalibration?: boolean
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/vad#processvad)

`processVAD`

Process audio samples for voice activity.

Copy

Ask AI

    await RunAnywhere.processVAD(audioSamples: number[]): Promise<VADResult>
    

### 

[​](https://docs.runanywhere.ai/react-native/vad#vadresult)

VADResult

Copy

Ask AI

    interface VADResult {
      /** Whether speech is detected */
      isSpeech: boolean
    
      /** Confidence score (0.0-1.0) */
      confidence: number
    
      /** Start time of speech segment (if detected) */
      startTime?: number
    
      /** End time of speech segment (if completed) */
      endTime?: number
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/vad#continuous-vad)

Continuous VAD

Continuous VAD is typically used via the Voice Agent pipeline (`startVoiceSession`), which handles VAD orchestration automatically. The methods below are for custom implementations.

Copy

Ask AI

    // Load the VAD model
    await RunAnywhere.loadVADModel(modelPath: string): Promise<void>
    
    // Start continuous VAD listening
    await RunAnywhere.startVAD(): Promise<void>
    
    // Stop continuous VAD listening
    await RunAnywhere.stopVAD(): Promise<void>
    
    // Set callback for speech activity events
    RunAnywhere.setVADSpeechActivityCallback(
      callback: (event: SpeechActivityEvent) => void
    ): void
    
    // Process a single audio frame (manual mode)
    await RunAnywhere.processVAD(audioSamples: number[]): Promise<VADResult>
    
    // Reset VAD state between sessions
    await RunAnywhere.resetVAD(): Promise<void>
    
    // Clean up VAD resources
    await RunAnywhere.cleanupVAD(): Promise<void>
    

### 

[​](https://docs.runanywhere.ai/react-native/vad#speechactivityevent)

SpeechActivityEvent

Copy

Ask AI

    interface SpeechActivityEvent {
      /** Event type */
      type: 'speechStarted' | 'speechEnded' | 'speechContinuing'
    
      /** Timestamp */
      timestamp: number
    
      /** Confidence */
      confidence: number
    
      /** Audio buffer (for speechEnded) */
      audioBuffer?: number[]
    }
    

[​](https://docs.runanywhere.ai/react-native/vad#examples)

Examples
----------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/vad#single-frame-processing)

Single Frame Processing

Copy

Ask AI

    // Process a single audio frame
    const result = await RunAnywhere.processVAD(audioFrame)
    
    console.log('Is speech:', result.isSpeech)
    console.log('Confidence:', result.confidence)
    

### 

[​](https://docs.runanywhere.ai/react-native/vad#continuous-listening)

Continuous Listening

VoiceListener.tsx

Copy

Ask AI

    import React, { useState, useEffect, useCallback } from 'react'
    import { View, Text, StyleSheet } from 'react-native'
    import { RunAnywhere, SpeechActivityEvent } from '@runanywhere/core'
    
    export function VoiceListener() {
      const [isSpeaking, setIsSpeaking] = useState(false)
      const [confidence, setConfidence] = useState(0)
    
      useEffect(() => {
        // Set up VAD callback
        RunAnywhere.setVADSpeechActivityCallback((event: SpeechActivityEvent) => {
          switch (event.type) {
            case 'speechStarted':
              setIsSpeaking(true)
              setConfidence(event.confidence)
              console.log('User started speaking')
              break
            case 'speechEnded':
              setIsSpeaking(false)
              console.log('User stopped speaking')
              // event.audioBuffer contains the speech audio
              break
            case 'speechContinuing':
              setConfidence(event.confidence)
              break
          }
        })
    
        // Start VAD
        RunAnywhere.startVAD()
    
        return () => {
          RunAnywhere.stopVAD()
        }
      }, [])
    
      return (
        <View style={styles.container}>
          <View style={[styles.indicator, isSpeaking && styles.speaking]} />
          <Text style={styles.status}>{isSpeaking ? 'Speaking...' : 'Listening...'}</Text>
          <Text style={styles.confidence}>Confidence: {(confidence * 100).toFixed(0)}%</Text>
        </View>
      )
    }
    
    const styles = StyleSheet.create({
      container: { alignItems: 'center', padding: 32 },
      indicator: {
        width: 80,
        height: 80,
        borderRadius: 40,
        backgroundColor: '#ddd',
        marginBottom: 16,
      },
      speaking: {
        backgroundColor: '#4caf50',
      },
      status: { fontSize: 18, fontWeight: '600' },
      confidence: { fontSize: 14, color: '#666', marginTop: 8 },
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/vad#vad-with-recording)

VAD with Recording

Copy

Ask AI

    import { RunAnywhere, SpeechActivityEvent } from '@runanywhere/core'
    
    class VoiceRecorder {
      private audioBuffer: number[] = []
      private isRecording = false
    
      async start() {
        RunAnywhere.setVADSpeechActivityCallback(this.handleVADEvent.bind(this))
        await RunAnywhere.startVAD()
      }
    
      private handleVADEvent(event: SpeechActivityEvent) {
        switch (event.type) {
          case 'speechStarted':
            this.isRecording = true
            this.audioBuffer = []
            console.log('Started recording')
            break
    
          case 'speechContinuing':
            // Audio is being accumulated by VAD
            break
    
          case 'speechEnded':
            this.isRecording = false
            if (event.audioBuffer) {
              this.audioBuffer = event.audioBuffer
              this.processRecording()
            }
            break
        }
      }
    
      private async processRecording() {
        // Transcribe the captured speech audio
        const samples = new Float32Array(this.audioBuffer)
        const result = await RunAnywhere.transcribeBuffer(samples, { sampleRate: 16000 })
        console.log('Transcription:', result.text)
      }
    
      async stop() {
        await RunAnywhere.stopVAD()
      }
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/vad#push-to-talk-alternative)

Push-to-Talk Alternative

PushToTalk.tsx

Copy

Ask AI

    import React, { useState, useCallback } from 'react'
    import { View, Pressable, Text, StyleSheet } from 'react-native'
    import { RunAnywhere } from '@runanywhere/core'
    
    export function PushToTalk({ onTranscription }: { onTranscription: (text: string) => void }) {
      const [isPressed, setIsPressed] = useState(false)
    
      const handlePressIn = useCallback(async () => {
        setIsPressed(true)
        // Start recording without VAD
        await RunAnywhere.startVAD()
      }, [])
    
      const handlePressOut = useCallback(async () => {
        setIsPressed(false)
        await RunAnywhere.stopVAD()
    
        // Process the recorded audio
        // (Implementation depends on how audio is captured)
      }, [])
    
      return (
        <Pressable
          onPressIn={handlePressIn}
          onPressOut={handlePressOut}
          style={[styles.button, isPressed && styles.pressed]}
        >
          <Text style={styles.text}>{isPressed ? '🎤 Recording...' : '🎤 Hold to Talk'}</Text>
        </Pressable>
      )
    }
    
    const styles = StyleSheet.create({
      button: {
        padding: 24,
        borderRadius: 12,
        backgroundColor: '#2196f3',
        alignItems: 'center',
      },
      pressed: {
        backgroundColor: '#f44336',
      },
      text: {
        color: '#fff',
        fontSize: 18,
        fontWeight: '600',
      },
    })
    

[​](https://docs.runanywhere.ai/react-native/vad#configuration-tips)

Configuration Tips
------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/vad#energy-threshold)

Energy Threshold

The `energyThreshold` determines how sensitive VAD is to speech:

Copy

Ask AI

    // More sensitive (picks up quiet speech, but more false positives)
    await RunAnywhere.initializeVAD({ energyThreshold: 0.3 })
    
    // Less sensitive (misses quiet speech, fewer false positives)
    await RunAnywhere.initializeVAD({ energyThreshold: 0.7 })
    
    // Balanced (default)
    await RunAnywhere.initializeVAD({ energyThreshold: 0.5 })
    

| Threshold | Sensitivity | Use Case |
| --- | --- | --- |
| 0.2-0.4 | High | Quiet environments |
| 0.4-0.6 | Medium | Normal conversation |
| 0.6-0.8 | Low | Noisy environments |

### 

[​](https://docs.runanywhere.ai/react-native/vad#auto-calibration)

Auto Calibration

Enable auto-calibration to adapt to ambient noise:

Copy

Ask AI

    await RunAnywhere.initializeVAD({
      autoCalibration: true,
      // Initial threshold will be adjusted based on background noise
    })
    

[​](https://docs.runanywhere.ai/react-native/vad#error-handling)

Error Handling
----------------------------------------------------------------------------------

Copy

Ask AI

    import { isSDKError, SDKErrorCode } from '@runanywhere/core'
    
    try {
      await RunAnywhere.processVAD(samples)
    } catch (error) {
      if (isSDKError(error)) {
        switch (error.code) {
          case SDKErrorCode.notInitialized:
            console.error('Initialize VAD first')
            break
          case SDKErrorCode.vadFailed:
            console.error('VAD processing failed')
            break
        }
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/vad#related)

Related
--------------------------------------------------------------------

[STT Transcribe\
--------------\
\
Speech-to-Text](https://docs.runanywhere.ai/react-native/stt/transcribe)
[Voice Agent\
-----------\
\
Full voice pipeline](https://docs.runanywhere.ai/react-native/voice-agent)
[STT Streaming\
-------------\
\
Real-time transcription](https://docs.runanywhere.ai/react-native/stt/stream)

[Voices](https://docs.runanywhere.ai/react-native/tts/voices)
[Voice Agent](https://docs.runanywhere.ai/react-native/voice-agent)

⌘I

---

# Streaming TTS - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/tts/stream#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

TTS (Text-to-Speech)

Streaming TTS

On this page

*   [Basic Usage](https://docs.runanywhere.ai/swift/tts/stream#basic-usage)
    
*   [Method Signature](https://docs.runanywhere.ai/swift/tts/stream#method-signature)
    
*   [Parameters](https://docs.runanywhere.ai/swift/tts/stream#parameters)
    
*   [Returns](https://docs.runanywhere.ai/swift/tts/stream#returns)
    
*   [Streaming Audio Player](https://docs.runanywhere.ai/swift/tts/stream#streaming-audio-player)
    
*   [SwiftUI Integration](https://docs.runanywhere.ai/swift/tts/stream#swiftui-integration)
    
*   [Sentence-by-Sentence Streaming](https://docs.runanywhere.ai/swift/tts/stream#sentence-by-sentence-streaming)
    
*   [Buffered Streaming](https://docs.runanywhere.ai/swift/tts/stream#buffered-streaming)
    
*   [Use Cases](https://docs.runanywhere.ai/swift/tts/stream#use-cases)
    
*   [Error Handling](https://docs.runanywhere.ai/swift/tts/stream#error-handling)
    
*   [Performance Tips](https://docs.runanywhere.ai/swift/tts/stream#performance-tips)
    

Streaming TTS generates audio in chunks, enabling playback to begin before the entire text is synthesized. This is ideal for long content where you want immediate audio feedback.

[​](https://docs.runanywhere.ai/swift/tts/stream#basic-usage)

Basic Usage
----------------------------------------------------------------------------

Copy

Ask AI

    try await RunAnywhere.synthesizeStream(
        "This is a long piece of text that will be synthesized in chunks...",
        options: TTSOptions(rate: 1.0),
        onAudioChunk: { chunk in
            // Play or buffer each audio chunk
            audioPlayer.enqueue(chunk)
        }
    )
    

[​](https://docs.runanywhere.ai/swift/tts/stream#method-signature)

Method Signature
--------------------------------------------------------------------------------------

Copy

Ask AI

    public static func synthesizeStream(
        _ text: String,
        options: TTSOptions = TTSOptions(),
        onAudioChunk: @escaping (Data) -> Void
    ) async throws -> TTSOutput
    

### 

[​](https://docs.runanywhere.ai/swift/tts/stream#parameters)

Parameters

| Parameter | Type | Description |
| --- | --- | --- |
| `text` | `String` | Text to synthesize |
| `options` | `TTSOptions` | Synthesis options |
| `onAudioChunk` | `(Data) -> Void` | Callback for each audio chunk |

### 

[​](https://docs.runanywhere.ai/swift/tts/stream#returns)

Returns

`TTSOutput` with the complete synthesized audio after streaming completes.

[​](https://docs.runanywhere.ai/swift/tts/stream#streaming-audio-player)

Streaming Audio Player
--------------------------------------------------------------------------------------------------

Build a player that handles streaming audio chunks:

Copy

Ask AI

    import AVFoundation
    
    class StreamingAudioPlayer {
        private var audioEngine: AVAudioEngine?
        private var playerNode: AVAudioPlayerNode?
        private var audioFormat: AVAudioFormat?
    
        func prepare(sampleRate: Double = 22050) throws {
            audioEngine = AVAudioEngine()
            playerNode = AVAudioPlayerNode()
    
            guard let engine = audioEngine, let player = playerNode else { return }
    
            audioFormat = AVAudioFormat(
                commonFormat: .pcmFormatFloat32,
                sampleRate: sampleRate,
                channels: 1,
                interleaved: false
            )
    
            engine.attach(player)
            engine.connect(player, to: engine.mainMixerNode, format: audioFormat)
    
            try engine.start()
            player.play()
        }
    
        func enqueue(_ audioData: Data) {
            guard let format = audioFormat,
                  let player = playerNode else { return }
    
            let frameCount = UInt32(audioData.count) / UInt32(MemoryLayout<Float>.size)
            guard let buffer = AVAudioPCMBuffer(pcmFormat: format, frameCapacity: frameCount) else { return }
    
            buffer.frameLength = frameCount
            audioData.withUnsafeBytes { ptr in
                if let address = ptr.baseAddress?.assumingMemoryBound(to: Float.self) {
                    buffer.floatChannelData?[0].update(from: address, count: Int(frameCount))
                }
            }
    
            player.scheduleBuffer(buffer)
        }
    
        func stop() {
            playerNode?.stop()
            audioEngine?.stop()
        }
    }
    

[​](https://docs.runanywhere.ai/swift/tts/stream#swiftui-integration)

SwiftUI Integration
--------------------------------------------------------------------------------------------

Copy

Ask AI

    struct StreamingTTSView: View {
        @State private var text = "Hello! This is a demonstration of streaming text-to-speech synthesis. The audio will begin playing before the entire text is processed, providing a more responsive experience for longer content."
        @State private var isSynthesizing = false
        @State private var progress: Double = 0
        @StateObject private var player = StreamingAudioPlayerWrapper()
    
        var body: some View {
            VStack(spacing: 20) {
                // Text input
                TextEditor(text: $text)
                    .frame(height: 150)
                    .border(Color.gray.opacity(0.3))
    
                // Progress indicator
                if isSynthesizing {
                    ProgressView(value: progress)
                        .progressViewStyle(.linear)
                }
    
                // Control button
                Button(action: synthesize) {
                    Label(
                        isSynthesizing ? "Synthesizing..." : "Speak",
                        systemImage: isSynthesizing ? "waveform" : "play.fill"
                    )
                    .frame(maxWidth: .infinity)
                    .padding()
                    .background(isSynthesizing ? Color.gray : Color.blue)
                    .foregroundColor(.white)
                    .cornerRadius(10)
                }
                .disabled(text.isEmpty || isSynthesizing)
            }
            .padding()
        }
    
        func synthesize() {
            isSynthesizing = true
            progress = 0
    
            Task {
                do {
                    try player.prepare()
    
                    var totalChunks = 0
                    var processedChunks = 0
    
                    // Estimate total chunks based on text length
                    totalChunks = max(1, text.count / 100)
    
                    let _ = try await RunAnywhere.synthesizeStream(
                        text,
                        options: TTSOptions(rate: 1.0),
                        onAudioChunk: { chunk in
                            player.enqueue(chunk)
                            processedChunks += 1
    
                            Task { @MainActor in
                                progress = min(1.0, Double(processedChunks) / Double(totalChunks))
                            }
                        }
                    )
    
                    await MainActor.run {
                        isSynthesizing = false
                        progress = 1.0
                    }
    
                } catch {
                    print("Synthesis failed: \(error)")
                    await MainActor.run {
                        isSynthesizing = false
                    }
                }
            }
        }
    }
    
    @MainActor
    class StreamingAudioPlayerWrapper: ObservableObject {
        private let player = StreamingAudioPlayer()
    
        func prepare() throws {
            try player.prepare()
        }
    
        func enqueue(_ data: Data) {
            player.enqueue(data)
        }
    
        func stop() {
            player.stop()
        }
    }
    

[​](https://docs.runanywhere.ai/swift/tts/stream#sentence-by-sentence-streaming)

Sentence-by-Sentence Streaming
------------------------------------------------------------------------------------------------------------------

For natural pauses, split text into sentences:

Copy

Ask AI

    func speakSentenceBySentence(_ text: String) async throws {
        let sentences = text.components(separatedBy: CharacterSet(charactersIn: ".!?"))
            .map { $0.trimmingCharacters(in: .whitespaces) }
            .filter { !$0.isEmpty }
    
        for sentence in sentences {
            let output = try await RunAnywhere.synthesize(
                sentence + ".",
                options: TTSOptions(rate: 1.0)
            )
    
            // Play and wait for completion
            let player = try AVAudioPlayer(data: output.audioData)
            player.play()
    
            // Wait for playback to complete
            try await Task.sleep(for: .seconds(output.duration))
        }
    }
    

[​](https://docs.runanywhere.ai/swift/tts/stream#buffered-streaming)

Buffered Streaming
------------------------------------------------------------------------------------------

Buffer chunks for smoother playback:

Copy

Ask AI

    class BufferedStreamingPlayer {
        private var buffer: [Data] = []
        private var isPlaying = false
        private let minimumBufferSize = 3  // Start playing after 3 chunks
    
        func addChunk(_ chunk: Data) {
            buffer.append(chunk)
    
            if buffer.count >= minimumBufferSize && !isPlaying {
                startPlayback()
            }
        }
    
        private func startPlayback() {
            isPlaying = true
    
            Task {
                while !buffer.isEmpty || isPlaying {
                    if let chunk = buffer.first {
                        buffer.removeFirst()
                        await playChunk(chunk)
                    } else {
                        try? await Task.sleep(for: .milliseconds(50))
                    }
                }
            }
        }
    
        private func playChunk(_ data: Data) async {
            // Play audio chunk
        }
    
        func finish() {
            isPlaying = false
        }
    }
    

[​](https://docs.runanywhere.ai/swift/tts/stream#use-cases)

Use Cases
------------------------------------------------------------------------

Audiobook reader

Stream long-form content with immediate playback.

Copy

Ask AI

    func readChapter(_ text: String) async throws {
        try await RunAnywhere.synthesizeStream(
            text,
            options: TTSOptions(rate: 0.9),
            onAudioChunk: { audioPlayer.enqueue($0) }
        )
    }
AI assistant voice

Stream LLM responses as they’re generated.

Copy

Ask AI

    let llmResult = try await RunAnywhere.generateStream(prompt)
    var accumulatedText = ""
    
    for try await token in llmResult.stream {
        accumulatedText += token
        
        // Synthesize complete sentences
        if token.contains(".") || token.contains("!") || token.contains("?") {
            try await RunAnywhere.synthesizeStream(accumulatedText) { chunk in
                player.enqueue(chunk)
            }
            accumulatedText = ""
        }
    }
Accessibility features

Read screen content aloud as user navigates.

Copy

Ask AI

    import SwiftUI
    
    struct AccessibleContentView: View {
        @State private var isSpeaking = false
        
        var body: some View {
            VStack {
                Text("Welcome to the app")
                    .accessibilityLabel("Welcome to the app")
                
                Button(isSpeaking ? "Stop Reading" : "Read Aloud") {
                    Task {
                        if isSpeaking {
                            await RunAnywhere.stopSpeaking()
                        } else {
                            isSpeaking = true
                            try? await RunAnywhere.synthesizeStream(
                                "Welcome to the app. Navigate using swipe gestures.",
                                options: TTSOptions(rate: 0.9),
                                onAudioChunk: { chunk in
                                    AudioPlayer.shared.enqueue(chunk)
                                }
                            )
                            isSpeaking = false
                        }
                    }
                }
                .accessibilityHint("Reads the current screen content aloud")
            }
        }
    }

[​](https://docs.runanywhere.ai/swift/tts/stream#error-handling)

Error Handling
----------------------------------------------------------------------------------

Copy

Ask AI

    do {
        try await RunAnywhere.synthesizeStream(
            text,
            options: TTSOptions(),
            onAudioChunk: { chunk in
                player.enqueue(chunk)
            }
        )
    } catch let error as SDKError {
        switch error.code {
        case .modelNotFound:
            print("Load a TTS voice first")
        case .processingFailed:
            print("Streaming synthesis failed")
        default:
            print("Error: \(error)")
        }
    
        // Stop player on error
        player.stop()
    }
    

[​](https://docs.runanywhere.ai/swift/tts/stream#performance-tips)

Performance Tips
--------------------------------------------------------------------------------------

**Buffer before playing**: Wait for a few chunks before starting playback to avoid stuttering.

**Match sample rates**: Ensure your audio player’s sample rate matches the TTS output (typically 22050 Hz).

[synthesize()\
------------\
\
Non-streaming synthesis →](https://docs.runanywhere.ai/swift/tts/synthesize)
[Voices\
------\
\
Available voices →](https://docs.runanywhere.ai/swift/tts/voices)

[synthesize()](https://docs.runanywhere.ai/swift/tts/synthesize)
[Voices](https://docs.runanywhere.ai/swift/tts/voices)

⌘I

---

# Voice Agent - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/swift/voice-agent#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

Swift SDK

Voice Agent

On this page

*   [Overview](https://docs.runanywhere.ai/swift/voice-agent#overview)
    
*   [Quick Start](https://docs.runanywhere.ai/swift/voice-agent#quick-start)
    
*   [Configuration](https://docs.runanywhere.ai/swift/voice-agent#configuration)
    
*   [VoiceAgentConfiguration](https://docs.runanywhere.ai/swift/voice-agent#voiceagentconfiguration)
    
*   [Initialize with Pre-loaded Models](https://docs.runanywhere.ai/swift/voice-agent#initialize-with-pre-loaded-models)
    
*   [VoiceAgentResult](https://docs.runanywhere.ai/swift/voice-agent#voiceagentresult)
    
*   [Component State](https://docs.runanywhere.ai/swift/voice-agent#component-state)
    
*   [ComponentLoadState](https://docs.runanywhere.ai/swift/voice-agent#componentloadstate)
    
*   [Individual Operations](https://docs.runanywhere.ai/swift/voice-agent#individual-operations)
    
*   [Complete Voice Assistant](https://docs.runanywhere.ai/swift/voice-agent#complete-voice-assistant)
    
*   [Conversation Context](https://docs.runanywhere.ai/swift/voice-agent#conversation-context)
    
*   [Error Handling](https://docs.runanywhere.ai/swift/voice-agent#error-handling)
    
*   [Cleanup](https://docs.runanywhere.ai/swift/voice-agent#cleanup)
    
*   [Best Practices](https://docs.runanywhere.ai/swift/voice-agent#best-practices)
    

The Voice Agent orchestrates the complete voice interaction pipeline: **VAD → STT → LLM → TTS**. It provides a unified API for building voice assistants, hands-free interfaces, and conversational AI.

[​](https://docs.runanywhere.ai/swift/voice-agent#overview)

Overview
-----------------------------------------------------------------------

The Voice Agent handles:

*   **VAD**: Detects when the user starts/stops speaking
*   **STT**: Transcribes speech to text
*   **LLM**: Generates AI responses
*   **TTS**: Synthesizes speech from responses

[​](https://docs.runanywhere.ai/swift/voice-agent#quick-start)

Quick Start
-----------------------------------------------------------------------------

Copy

Ask AI

    // Initialize the voice agent
    try await RunAnywhere.initializeVoiceAgent(
        VoiceAgentConfiguration(
            sttModelId: "whisper-base-onnx",
            llmModelId: "llama-3.2-1b-instruct-q4",
            ttsVoice: "piper-en-us-amy"
        )
    )
    
    // Process a voice turn
    let audioData: Data = // Recorded user speech
    let result = try await RunAnywhere.processVoiceTurn(audioData)
    
    print("User said: \(result.transcription ?? "")")
    print("AI response: \(result.response ?? "")")
    
    // Play the synthesized audio
    if let audio = result.synthesizedAudio {
        playAudio(audio)
    }
    

[​](https://docs.runanywhere.ai/swift/voice-agent#configuration)

Configuration
---------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/swift/voice-agent#voiceagentconfiguration)

VoiceAgentConfiguration

Copy

Ask AI

    public struct VoiceAgentConfiguration: Sendable {
        public let sttModelId: String?       // STT model (e.g., "whisper-base-onnx")
        public let llmModelId: String?       // LLM model (e.g., "llama-3.2-1b-instruct-q4")
        public let ttsVoice: String?         // TTS voice (e.g., "piper-en-us-amy")
        public let vadSampleRate: Int        // VAD sample rate (default: 16000)
        public let vadFrameLength: Float     // VAD frame length (default: 0.032)
        public let vadEnergyThreshold: Float // VAD sensitivity (default: 0.5)
    
        public init(
            sttModelId: String? = nil,
            llmModelId: String? = nil,
            ttsVoice: String? = nil,
            vadSampleRate: Int = 16000,
            vadFrameLength: Float = 0.032,
            vadEnergyThreshold: Float = 0.5
        )
    }
    

### 

[​](https://docs.runanywhere.ai/swift/voice-agent#initialize-with-pre-loaded-models)

Initialize with Pre-loaded Models

If you’ve already loaded models, use:

Copy

Ask AI

    // Load models separately
    try await RunAnywhere.loadSTTModel("whisper-base-onnx")
    try await RunAnywhere.loadModel("llama-3.2-1b-instruct-q4")
    try await RunAnywhere.loadTTSVoice("piper-en-us-amy")
    
    // Initialize voice agent with loaded models
    try await RunAnywhere.initializeVoiceAgentWithLoadedModels()
    

[​](https://docs.runanywhere.ai/swift/voice-agent#voiceagentresult)

VoiceAgentResult
---------------------------------------------------------------------------------------

Copy

Ask AI

    public struct VoiceAgentResult: Sendable {
        public let speechDetected: Bool       // Was speech detected?
        public let transcription: String?     // What the user said
        public let response: String?          // AI response text
        public let synthesizedAudio: Data?    // Audio of response
    }
    

[​](https://docs.runanywhere.ai/swift/voice-agent#component-state)

Component State
-------------------------------------------------------------------------------------

Check readiness of individual components:

Copy

Ask AI

    // Check all components
    let states = await RunAnywhere.getVoiceAgentComponentStates()
    print("STT: \(states.stt)")
    print("LLM: \(states.llm)")
    print("TTS: \(states.tts)")
    
    // Check if everything is ready
    if await RunAnywhere.areAllVoiceComponentsReady {
        // Ready for voice interaction
    }
    
    // Check if voice agent is initialized
    let isReady = await RunAnywhere.isVoiceAgentReady
    

### 

[​](https://docs.runanywhere.ai/swift/voice-agent#componentloadstate)

ComponentLoadState

Copy

Ask AI

    public enum ComponentLoadState: Sendable {
        case notLoaded
        case loading
        case loaded(modelId: String)
        case failed(Error)
    }
    

[​](https://docs.runanywhere.ai/swift/voice-agent#individual-operations)

Individual Operations
-------------------------------------------------------------------------------------------------

Use components separately when needed:

Copy

Ask AI

    // Just transcribe
    let transcription = try await RunAnywhere.voiceAgentTranscribe(audioData)
    
    // Just generate response
    let response = try await RunAnywhere.voiceAgentGenerateResponse(transcription)
    
    // Just synthesize speech
    let audio = try await RunAnywhere.voiceAgentSynthesizeSpeech(response)
    

[​](https://docs.runanywhere.ai/swift/voice-agent#complete-voice-assistant)

Complete Voice Assistant
-------------------------------------------------------------------------------------------------------

Copy

Ask AI

    import SwiftUI
    import AVFoundation
    
    @Observable
    class VoiceAssistant {
        var isListening = false
        var isProcessing = false
        var transcription = ""
        var response = ""
        var conversationHistory: [(role: String, text: String)] = []
    
        private var audioEngine: AVAudioEngine?
        private var audioRecorder: AVAudioRecorder?
        private var audioPlayer: AVAudioPlayer?
    
        func initialize() async throws {
            // Initialize voice agent
            try await RunAnywhere.initializeVoiceAgent(
                VoiceAgentConfiguration(
                    sttModelId: "whisper-base-onnx",
                    llmModelId: "llama-3.2-1b-instruct-q4",
                    ttsVoice: "piper-en-us-amy",
                    vadEnergyThreshold: 0.5
                )
            )
        }
    
        func startListening() async throws {
            let session = AVAudioSession.sharedInstance()
            try session.setCategory(.playAndRecord, mode: .default)
            try session.setActive(true)
    
            let url = FileManager.default.temporaryDirectory.appendingPathComponent("voice.wav")
    
            let settings: [String: Any] = [\
                AVFormatIDKey: Int(kAudioFormatLinearPCM),\
                AVSampleRateKey: 16000,\
                AVNumberOfChannelsKey: 1,\
                AVLinearPCMBitDepthKey: 16,\
                AVLinearPCMIsFloatKey: false\
            ]
    
            audioRecorder = try AVAudioRecorder(url: url, settings: settings)
            audioRecorder?.record()
    
            await MainActor.run { isListening = true }
        }
    
        func stopAndProcess() async throws {
            guard let recorder = audioRecorder else { return }
    
            recorder.stop()
            await MainActor.run {
                isListening = false
                isProcessing = true
            }
    
            let audioData = try Data(contentsOf: recorder.url)
    
            // Process voice turn
            let result = try await RunAnywhere.processVoiceTurn(audioData)
    
            await MainActor.run {
                transcription = result.transcription ?? ""
                response = result.response ?? ""
    
                if !transcription.isEmpty {
                    conversationHistory.append((role: "user", text: transcription))
                }
                if !response.isEmpty {
                    conversationHistory.append((role: "assistant", text: response))
                }
    
                isProcessing = false
            }
    
            // Play response
            if let audio = result.synthesizedAudio {
                audioPlayer = try AVAudioPlayer(data: audio)
                audioPlayer?.play()
            }
        }
    
        func cleanup() async {
            await RunAnywhere.cleanupVoiceAgent()
        }
    }
    
    struct VoiceAssistantView: View {
        @State private var assistant = VoiceAssistant()
        @State private var isInitialized = false
        @State private var error: String?
    
        var body: some View {
            VStack(spacing: 20) {
                // Conversation history
                ScrollView {
                    LazyVStack(alignment: .leading, spacing: 12) {
                        ForEach(Array(assistant.conversationHistory.enumerated()), id: \.offset) { _, message in
                            HStack {
                                if message.role == "user" { Spacer() }
    
                                VStack(alignment: message.role == "user" ? .trailing : .leading) {
                                    Text(message.role == "user" ? "You" : "Assistant")
                                        .font(.caption)
                                        .foregroundColor(.secondary)
    
                                    Text(message.text)
                                        .padding()
                                        .background(message.role == "user" ? Color.blue : Color.gray.opacity(0.2))
                                        .foregroundColor(message.role == "user" ? .white : .primary)
                                        .cornerRadius(16)
                                }
    
                                if message.role == "assistant" { Spacer() }
                            }
                        }
                    }
                    .padding()
                }
    
                // Voice button
                Button(action: toggleVoice) {
                    ZStack {
                        Circle()
                            .fill(buttonColor)
                            .frame(width: 80, height: 80)
    
                        if assistant.isProcessing {
                            ProgressView()
                                .tint(.white)
                        } else {
                            Image(systemName: assistant.isListening ? "stop.fill" : "mic.fill")
                                .font(.title)
                                .foregroundColor(.white)
                        }
                    }
                }
                .disabled(!isInitialized || assistant.isProcessing)
    
                Text(statusText)
                    .font(.caption)
                    .foregroundColor(.secondary)
            }
            .task {
                do {
                    try await assistant.initialize()
                    isInitialized = true
                } catch {
                    self.error = error.localizedDescription
                }
            }
            .alert("Error", isPresented: .constant(error != nil)) {
                Button("OK") { error = nil }
            } message: {
                Text(error ?? "")
            }
        }
    
        var buttonColor: Color {
            if !isInitialized { return .gray }
            if assistant.isProcessing { return .orange }
            if assistant.isListening { return .red }
            return .blue
        }
    
        var statusText: String {
            if !isInitialized { return "Initializing..." }
            if assistant.isProcessing { return "Processing..." }
            if assistant.isListening { return "Listening... tap to stop" }
            return "Tap to speak"
        }
    
        func toggleVoice() {
            Task {
                if assistant.isListening {
                    try? await assistant.stopAndProcess()
                } else {
                    try? await assistant.startListening()
                }
            }
        }
    }
    

[​](https://docs.runanywhere.ai/swift/voice-agent#conversation-context)

Conversation Context
-----------------------------------------------------------------------------------------------

Maintain conversation history for context-aware responses:

Copy

Ask AI

    class ConversationalVoiceAgent {
        private var conversationContext: [Message] = []
    
        struct Message {
            let role: String
            let content: String
        }
    
        func processWithContext(_ audioData: Data) async throws -> VoiceAgentResult {
            // Transcribe
            let transcription = try await RunAnywhere.voiceAgentTranscribe(audioData)
    
            // Add to context
            conversationContext.append(Message(role: "user", content: transcription))
    
            // Build context prompt
            let contextPrompt = conversationContext
                .suffix(10)  // Last 10 messages
                .map { "\($0.role): \($0.content)" }
                .joined(separator: "\n")
    
            // Generate response with context
            let response = try await RunAnywhere.voiceAgentGenerateResponse(contextPrompt)
    
            // Add response to context
            conversationContext.append(Message(role: "assistant", content: response))
    
            // Synthesize
            let audio = try await RunAnywhere.voiceAgentSynthesizeSpeech(response)
    
            return VoiceAgentResult(
                speechDetected: true,
                transcription: transcription,
                response: response,
                synthesizedAudio: audio
            )
        }
    
        func clearContext() {
            conversationContext.removeAll()
        }
    }
    

[​](https://docs.runanywhere.ai/swift/voice-agent#error-handling)

Error Handling
-----------------------------------------------------------------------------------

Copy

Ask AI

    do {
        try await RunAnywhere.initializeVoiceAgent(config)
        let result = try await RunAnywhere.processVoiceTurn(audioData)
    } catch let error as SDKError {
        switch error.code {
        case .notInitialized:
            print("Initialize voice agent first")
        case .modelNotFound:
            print("Required model not loaded")
        case .processingFailed:
            print("Processing failed: \(error.message)")
        default:
            print("Error: \(error)")
        }
    }
    

[​](https://docs.runanywhere.ai/swift/voice-agent#cleanup)

Cleanup
---------------------------------------------------------------------

Always clean up when done:

Copy

Ask AI

    await RunAnywhere.cleanupVoiceAgent()
    

[​](https://docs.runanywhere.ai/swift/voice-agent#best-practices)

Best Practices
-----------------------------------------------------------------------------------

Pre-load models

Load all models during app initialization to avoid delays during conversation.

Handle interruptions

Properly handle audio session interruptions (phone calls, etc.).

Visual feedback

Show clear visual indicators for listening, processing, and speaking states.

Error recovery

Implement graceful error recovery to maintain conversation flow.

[VAD\
---\
\
Voice Activity Detection →](https://docs.runanywhere.ai/swift/vad)
[STT\
---\
\
Speech-to-Text →](https://docs.runanywhere.ai/swift/stt/transcribe)

[Voice Activity Detection](https://docs.runanywhere.ai/swift/vad)
[Configuration](https://docs.runanywhere.ai/swift/configuration)

⌘I

---

# Voice Agent - RunAnywhere Documentation

[Skip to main content](https://docs.runanywhere.ai/react-native/voice-agent#content-area)

[RunAnywhere Documentation home page![light logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)![dark logo](https://mintcdn.com/runanywhere/LBew9r1yliJTxAIH/logo-gradient.svg?fit=max&auto=format&n=LBew9r1yliJTxAIH&q=85&s=8321e635ba8217e2714200582a9b09a8)](https://docs.runanywhere.ai/)

Search...

⌘KAsk AI

Search...

Navigation

React Native SDK

Voice Agent

On this page

*   [Overview](https://docs.runanywhere.ai/react-native/voice-agent#overview)
    
*   [Basic Usage](https://docs.runanywhere.ai/react-native/voice-agent#basic-usage)
    
*   [Setup](https://docs.runanywhere.ai/react-native/voice-agent#setup)
    
*   [1\. Download Required Models](https://docs.runanywhere.ai/react-native/voice-agent#1-download-required-models)
    
*   [2\. Initialize Voice Agent](https://docs.runanywhere.ai/react-native/voice-agent#2-initialize-voice-agent)
    
*   [API Reference](https://docs.runanywhere.ai/react-native/voice-agent#api-reference)
    
*   [initializeVoiceAgent](https://docs.runanywhere.ai/react-native/voice-agent#initializevoiceagent)
    
*   [VoiceAgentConfig](https://docs.runanywhere.ai/react-native/voice-agent#voiceagentconfig)
    
*   [processVoiceTurn](https://docs.runanywhere.ai/react-native/voice-agent#processvoiceturn)
    
*   [VoiceTurnResult](https://docs.runanywhere.ai/react-native/voice-agent#voiceturnresult)
    
*   [startVoiceSession](https://docs.runanywhere.ai/react-native/voice-agent#startvoicesession)
    
*   [VoiceSessionConfig](https://docs.runanywhere.ai/react-native/voice-agent#voicesessionconfig)
    
*   [VoiceSessionEvent](https://docs.runanywhere.ai/react-native/voice-agent#voicesessionevent)
    
*   [Examples](https://docs.runanywhere.ai/react-native/voice-agent#examples)
    
*   [Complete Voice Assistant](https://docs.runanywhere.ai/react-native/voice-agent#complete-voice-assistant)
    
*   [Single Turn Processing](https://docs.runanywhere.ai/react-native/voice-agent#single-turn-processing)
    
*   [Custom Voice Pipeline](https://docs.runanywhere.ai/react-native/voice-agent#custom-voice-pipeline)
    
*   [Performance Optimization](https://docs.runanywhere.ai/react-native/voice-agent#performance-optimization)
    
*   [Latency Breakdown](https://docs.runanywhere.ai/react-native/voice-agent#latency-breakdown)
    
*   [Tips for Lower Latency](https://docs.runanywhere.ai/react-native/voice-agent#tips-for-lower-latency)
    
*   [Error Handling](https://docs.runanywhere.ai/react-native/voice-agent#error-handling)
    
*   [Related](https://docs.runanywhere.ai/react-native/voice-agent#related)
    

[​](https://docs.runanywhere.ai/react-native/voice-agent#overview)

Overview
------------------------------------------------------------------------------

The Voice Agent provides a complete voice conversation pipeline that orchestrates VAD → STT → LLM → TTS in a seamless flow. Use it to build voice assistants, hands-free interfaces, and conversational AI applications.

[​](https://docs.runanywhere.ai/react-native/voice-agent#basic-usage)

Basic Usage
------------------------------------------------------------------------------------

Copy

Ask AI

    import { RunAnywhere } from '@runanywhere/core'
    
    // Initialize voice agent
    await RunAnywhere.initializeVoiceAgent({
      llmModelId: 'smollm2-360m',
      sttModelId: 'whisper-tiny-en',
      ttsModelId: 'piper-en-lessac',
      systemPrompt: 'You are a helpful voice assistant. Keep responses brief.',
    })
    
    // Process a complete voice turn
    const result = await RunAnywhere.processVoiceTurn(audioData)
    
    console.log('User said:', result.userTranscript)
    console.log('AI response:', result.assistantResponse)
    // result.audio contains the spoken response
    

[​](https://docs.runanywhere.ai/react-native/voice-agent#setup)

Setup
------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#1-download-required-models)

1\. Download Required Models

Copy

Ask AI

    import { RunAnywhere, SDKEnvironment, ModelCategory } from '@runanywhere/core'
    import { LlamaCPP } from '@runanywhere/llamacpp'
    import { ONNX, ModelArtifactType } from '@runanywhere/onnx'
    
    // Initialize SDK
    await RunAnywhere.initialize({ environment: SDKEnvironment.Development })
    
    // Register backends
    LlamaCPP.register()
    ONNX.register()
    
    // Add models
    await LlamaCPP.addModel({
      id: 'smollm2-360m',
      name: 'SmolLM2 360M',
      url: 'https://huggingface.co/.../SmolLM2-360M.Q8_0.gguf',
      memoryRequirement: 500_000_000,
    })
    
    await ONNX.addModel({
      id: 'whisper-tiny-en',
      name: 'Whisper Tiny English',
      url: 'https://github.com/.../sherpa-onnx-whisper-tiny.en.tar.gz',
      modality: ModelCategory.SpeechRecognition,
      artifactType: ModelArtifactType.TarGzArchive,
      memoryRequirement: 75_000_000,
    })
    
    await ONNX.addModel({
      id: 'piper-en-lessac',
      name: 'Piper English',
      url: 'https://github.com/.../vits-piper-en_US-lessac-medium.tar.gz',
      modality: ModelCategory.SpeechSynthesis,
      artifactType: ModelArtifactType.TarGzArchive,
      memoryRequirement: 65_000_000,
    })
    
    // Download all models
    await Promise.all([\
      RunAnywhere.downloadModel('smollm2-360m'),\
      RunAnywhere.downloadModel('whisper-tiny-en'),\
      RunAnywhere.downloadModel('piper-en-lessac'),\
    ])
    

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#2-initialize-voice-agent)

2\. Initialize Voice Agent

Copy

Ask AI

    await RunAnywhere.initializeVoiceAgent({
      llmModelId: 'smollm2-360m',
      sttModelId: 'whisper-tiny-en',
      ttsModelId: 'piper-en-lessac',
      systemPrompt: 'You are a helpful voice assistant. Be concise and friendly.',
      generationOptions: {
        maxTokens: 150,
        temperature: 0.7,
      },
    })
    

[​](https://docs.runanywhere.ai/react-native/voice-agent#api-reference)

API Reference
----------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#initializevoiceagent)

`initializeVoiceAgent`

Initialize the voice agent pipeline.

Copy

Ask AI

    await RunAnywhere.initializeVoiceAgent(
      config: VoiceAgentConfig
    ): Promise<boolean>
    

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#voiceagentconfig)

VoiceAgentConfig

Copy

Ask AI

    interface VoiceAgentConfig {
      /** LLM model ID (must be downloaded) */
      llmModelId: string
    
      /** STT model ID (must be downloaded) */
      sttModelId: string
    
      /** TTS model ID (must be downloaded) */
      ttsModelId: string
    
      /** System prompt for LLM */
      systemPrompt?: string
    
      /** Generation options */
      generationOptions?: GenerationOptions
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#processvoiceturn)

`processVoiceTurn`

Process a complete voice turn (STT → LLM → TTS).

Copy

Ask AI

    await RunAnywhere.processVoiceTurn(
      audioData: string
    ): Promise<VoiceTurnResult>
    

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#voiceturnresult)

VoiceTurnResult

Copy

Ask AI

    interface VoiceTurnResult {
      /** Transcribed user speech */
      userTranscript: string
    
      /** LLM-generated response */
      assistantResponse: string
    
      /** Synthesized audio (base64 float32 PCM) */
      audio: string
    
      /** Performance metrics */
      metrics: VoiceAgentMetrics
    }
    
    interface VoiceAgentMetrics {
      /** STT latency in ms */
      sttLatencyMs: number
    
      /** LLM latency in ms */
      llmLatencyMs: number
    
      /** TTS latency in ms */
      ttsLatencyMs: number
    
      /** Total turn latency in ms */
      totalLatencyMs: number
    
      /** Tokens generated */
      tokensGenerated: number
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#startvoicesession)

`startVoiceSession`

Start an interactive voice session with automatic VAD.

Copy

Ask AI

    await RunAnywhere.startVoiceSession(
      config: VoiceSessionConfig,
      callback: (event: VoiceSessionEvent) => void
    ): Promise<VoiceSessionHandle>
    

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#voicesessionconfig)

VoiceSessionConfig

Copy

Ask AI

    interface VoiceSessionConfig {
      /** Seconds of silence before processing speech (default: 1.5) */
      silenceDuration?: number
    
      /** Minimum audio level to detect speech, 0.0-1.0 (default: 0.1) */
      speechThreshold?: number
    
      /** Auto-play TTS audio (default: true) */
      autoPlayTTS?: boolean
    
      /** Auto-resume listening after TTS finishes (default: true) */
      continuousMode?: boolean
    
      /** Language code (default: 'en') */
      language?: string
    
      /** System prompt for LLM */
      systemPrompt?: string
    
      /** Event callback */
      onEvent?: (event: VoiceSessionEvent) => void
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#voicesessionevent)

VoiceSessionEvent

Copy

Ask AI

    interface VoiceSessionEvent {
      type:
        | 'started'
        | 'listening'
        | 'speechStarted'
        | 'speechEnded'
        | 'processing'
        | 'transcribed'
        | 'responded'
        | 'speaking'
        | 'turnCompleted'
        | 'stopped'
        | 'error'
    
      /** Timestamp */
      timestamp: number
    
      /** Audio input level, 0.0-1.0 (for 'listening' events) */
      audioLevel?: number
    
      /** Transcribed text (for 'transcribed' and 'turnCompleted') */
      transcription?: string
    
      /** LLM response text (for 'responded' and 'turnCompleted') */
      response?: string
    
      /** TTS audio data (for 'turnCompleted') */
      audio?: string
    
      /** Error message (for 'error' events) */
      error?: string
    }
    

[​](https://docs.runanywhere.ai/react-native/voice-agent#examples)

Examples
------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#complete-voice-assistant)

Complete Voice Assistant

VoiceAssistant.tsx

Copy

Ask AI

    import React, { useState, useCallback, useEffect } from 'react'
    import { View, Text, Button, StyleSheet } from 'react-native'
    import { RunAnywhere, VoiceSessionEvent, VoiceSessionHandle } from '@runanywhere/core'
    
    export function VoiceAssistant() {
      const [isActive, setIsActive] = useState(false)
      const [status, setStatus] = useState('Ready')
      const [transcript, setTranscript] = useState('')
      const [response, setResponse] = useState('')
      const sessionRef = React.useRef<VoiceSessionHandle | null>(null)
    
      const handleEvent = useCallback((event: VoiceSessionEvent) => {
        switch (event.type) {
          case 'started':
            setStatus('Session started')
            break
          case 'listening':
            setStatus('Listening...')
            break
          case 'speechStarted':
            setStatus('Hearing you...')
            break
          case 'processing':
            setStatus('Processing...')
            break
          case 'transcribed':
            setTranscript(event.transcription || '')
            setStatus('Thinking...')
            break
          case 'responded':
            setResponse(event.response || '')
            setStatus('Speaking...')
            break
          case 'turnCompleted':
            setStatus('Listening...')
            break
          case 'error':
            setStatus('Error: ' + event.error)
            break
        }
      }, [])
    
      const startSession = useCallback(async () => {
        setIsActive(true)
        setTranscript('')
        setResponse('')
    
        sessionRef.current = await RunAnywhere.startVoiceSession({
          systemPrompt: 'You are a helpful voice assistant.',
          continuousMode: true,
          onEvent: handleEvent,
        })
      }, [handleEvent])
    
      const stopSession = useCallback(async () => {
        if (sessionRef.current) {
          await sessionRef.current.stop()
          sessionRef.current = null
        }
        setIsActive(false)
        setStatus('Ready')
      }, [])
    
      return (
        <View style={styles.container}>
          <View style={[styles.statusBadge, isActive && styles.active]}>
            <Text style={styles.statusText}>{status}</Text>
          </View>
    
          {transcript && (
            <View style={styles.bubble}>
              <Text style={styles.bubbleLabel}>You</Text>
              <Text style={styles.bubbleText}>{transcript}</Text>
            </View>
          )}
    
          {response && (
            <View style={[styles.bubble, styles.assistantBubble]}>
              <Text style={styles.bubbleLabel}>Assistant</Text>
              <Text style={styles.bubbleText}>{response}</Text>
            </View>
          )}
    
          <Button
            title={isActive ? 'Stop' : 'Start Voice Assistant'}
            onPress={isActive ? stopSession : startSession}
          />
        </View>
      )
    }
    
    const styles = StyleSheet.create({
      container: { flex: 1, padding: 16 },
      statusBadge: {
        alignSelf: 'center',
        paddingHorizontal: 16,
        paddingVertical: 8,
        borderRadius: 20,
        backgroundColor: '#e0e0e0',
        marginBottom: 24,
      },
      active: { backgroundColor: '#4caf50' },
      statusText: { fontWeight: '600', color: '#333' },
      bubble: {
        padding: 12,
        borderRadius: 12,
        backgroundColor: '#f5f5f5',
        marginBottom: 12,
      },
      assistantBubble: { backgroundColor: '#e3f2fd' },
      bubbleLabel: { fontSize: 12, color: '#666', marginBottom: 4 },
      bubbleText: { fontSize: 16 },
    })
    

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#single-turn-processing)

Single Turn Processing

Copy

Ask AI

    // For push-to-talk or single-turn interactions
    async function processSingleTurn(audioBase64: string) {
      const result = await RunAnywhere.processVoiceTurn(audioBase64)
    
      console.log('=== Voice Turn Results ===')
      console.log('User:', result.userTranscript)
      console.log('Assistant:', result.assistantResponse)
      console.log('')
      console.log('=== Metrics ===')
      console.log('STT:', result.metrics.sttLatencyMs, 'ms')
      console.log('LLM:', result.metrics.llmLatencyMs, 'ms')
      console.log('TTS:', result.metrics.ttsLatencyMs, 'ms')
      console.log('Total:', result.metrics.totalLatencyMs, 'ms')
    
      // Play the response audio
      playAudio(result.audio)
    
      return result
    }
    

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#custom-voice-pipeline)

Custom Voice Pipeline

Build a custom pipeline for more control:

Copy

Ask AI

    class CustomVoiceAgent {
      private conversationHistory: { role: string; content: string }[] = []
    
      async processTurn(audioSamples: Float32Array): Promise<void> {
        // 1. Transcribe user speech
        const sttResult = await RunAnywhere.transcribeBuffer(audioSamples, {
          language: 'en',
        })
        console.log('User:', sttResult.text)
    
        // 2. Add to conversation history
        this.conversationHistory.push({
          role: 'user',
          content: sttResult.text,
        })
    
        // 3. Generate LLM response with history
        const prompt = this.buildPrompt()
        const llmResult = await RunAnywhere.generate(prompt, {
          maxTokens: 150,
          temperature: 0.7,
          systemPrompt: 'You are a helpful assistant.',
        })
        console.log('Assistant:', llmResult.text)
    
        // 4. Add response to history
        this.conversationHistory.push({
          role: 'assistant',
          content: llmResult.text,
        })
    
        // 5. Synthesize and play response
        const ttsResult = await RunAnywhere.synthesize(llmResult.text)
        await playAudio(ttsResult.audio, ttsResult.sampleRate)
      }
    
      private buildPrompt(): string {
        return (
          this.conversationHistory
            .map((m) => `${m.role === 'user' ? 'User' : 'Assistant'}: ${m.content}`)
            .join('\n') + '\nAssistant:'
        )
      }
    
      clearHistory() {
        this.conversationHistory = []
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/voice-agent#performance-optimization)

Performance Optimization
--------------------------------------------------------------------------------------------------------------

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#latency-breakdown)

Latency Breakdown

| Stage | Typical Latency | Optimization |
| --- | --- | --- |
| VAD | < 50ms | Use appropriate frame size |
| STT | 100-500ms | Use smaller models |
| LLM | 200-2000ms | Use smaller models, limit tokens |
| TTS | 100-300ms | Use streaming |
| **Total** | **400-3000ms** |     |

### 

[​](https://docs.runanywhere.ai/react-native/voice-agent#tips-for-lower-latency)

Tips for Lower Latency

Use smaller models for faster response times. A 360M LLM with Whisper Tiny can achieve sub-second voice turns.

Copy

Ask AI

    // Optimized for speed
    await RunAnywhere.initializeVoiceAgent({
      llmModelId: 'smollm2-360m', // Small, fast model
      sttModelId: 'whisper-tiny-en', // Fastest STT
      ttsModelId: 'piper-en-lessac', // Neural TTS
      generationOptions: {
        maxTokens: 100, // Shorter responses
        temperature: 0.5, // Faster sampling
      },
    })
    

[​](https://docs.runanywhere.ai/react-native/voice-agent#error-handling)

Error Handling
------------------------------------------------------------------------------------------

Copy

Ask AI

    try {
      const session = await RunAnywhere.startVoiceSession({
        ...config,
        onEvent: (event) => {
          if (event.type === 'error') {
            handleError(event.error)
          }
        },
      })
    } catch (error) {
      if (isSDKError(error)) {
        switch (error.code) {
          case SDKErrorCode.modelNotFound:
            console.error('Download required models first')
            break
          case SDKErrorCode.voiceAgentFailed:
            console.error('Voice agent failed:', error.message)
            break
        }
      }
    }
    

[​](https://docs.runanywhere.ai/react-native/voice-agent#related)

Related
----------------------------------------------------------------------------

[VAD\
---\
\
Voice Activity Detection](https://docs.runanywhere.ai/react-native/vad)
[STT\
---\
\
Speech-to-Text](https://docs.runanywhere.ai/react-native/stt/transcribe)
[TTS\
---\
\
Text-to-Speech](https://docs.runanywhere.ai/react-native/tts/synthesize)
[LLM\
---\
\
Text Generation](https://docs.runanywhere.ai/react-native/llm/generate)

[Voice Activity Detection](https://docs.runanywhere.ai/react-native/vad)
[Configuration](https://docs.runanywhere.ai/react-native/configuration)

⌘I

---