API Reference

Main entry point for the CompositeVoice SDK.

Remarks

Re-exports all public classes, interfaces, types, and utilities from the SDK. This is the primary module consumers import from when using the SDK.

The SDK provides a 5-role audio pipeline with pluggable providers for Audio Input, Speech-to-Text (STT), Large Language Models (LLM), Text-to-Speech (TTS), and Audio Output. Providers declare their roles via a roles property, and multi-role providers (e.g., NativeSTT covering both input and stt) are supported via array-based config.

Pipeline: [InputProvider] → InputQueue → [STT] → [LLM] → [TTS] → OutputQueue → [OutputProvider]

It features an event-driven architecture, conversation history management, eager LLM pipeline support, configurable turn-taking strategies, audio buffering between pipeline stages, format detection with header caching, and WebSocket reconnection with exponential backoff.

Examples

import {
  CompositeVoice,
  NativeSTT,
  AnthropicLLM,
  NativeTTS,
} from '@lukeocodes/composite-voice';

const voice = new CompositeVoice({
  providers: [
    new NativeSTT(),
    new AnthropicLLM({ proxyUrl: '/api/proxy/anthropic', model: 'claude-haiku-4-5' }),
    new NativeTTS(),
  ],
});

voice.on('agent.stateChange', ({ state }) => console.log('State:', state));
voice.on('transcription.speechFinal', ({ text }) => console.log('User:', text));
voice.on('tts.audio', ({ audio }) => console.log('Audio chunk received'));

await voice.initialize();
await voice.startListening();

import { BaseLLMProvider } from '@lukeocodes/composite-voice';

class MyCustomLLM extends BaseLLMProvider {
  // ... implement abstract methods
}

See

• CompositeVoice for the main orchestrator class
• GitHub repository for full documentation

Browse by category

• Classes — Concrete class implementations — providers, orchestrators, and core SDK components. (30)
• Abstract Classes — Base classes for building custom STT, LLM, and TTS providers. (4)
• Interfaces — Configuration types, provider contracts, and event structures. (58)
• Types — Type aliases, union types, and utility types. (30)
• Functions — Utility functions for audio processing, logging, and more. (18)
• Errors — Error classes with machine-readable codes and recovery flags. (11)
• Enumerations — Enum types used across the SDK. (1)