Austack Client SDK for Node.js/TypeScript

A TypeScript/JavaScript client SDK for Austack conversational AI, providing WebSocket-based audio communication with built-in audio processing and conversation management.

Installation

npm install @austack/client

Features

• 🎙️ Audio Input/Output: Browser-compatible audio capture and playback
• 🔊 Real-time Communication: WebSocket-based bidirectional audio streaming
• 🎯 Voice Activity Detection: Intelligent speech detection and silence handling
• 🛠️ TypeScript Support: Full type definitions included
• 🔄 Interrupt Handling: Smart audio interruption for natural conversations
• 📦 Dual Module Support: Works with both CommonJS and ES modules

Quick Start

import { ConversationManager } from '@austack/client';

const conversation = new ConversationManager(
  'ws://localhost:8000/ws',
  (state) => {
    console.log('Conversation state:', state);
  }
);

// Start the conversation
await conversation.startConversation();

// Stop the conversation
conversation.stopConversation();

// Clean up resources
conversation.cleanup();

API Reference

ConversationManager

The main class for managing audio conversations.

class ConversationManager {
  constructor(
    websocketUrl: string,
    stateChangeCallback?: (state: ConversationState) => void
  )

  async startConversation(): Promise<void>
  stopConversation(): void
  sendMessage(message: object): void
  cleanup(): void
  getState(): ConversationState
  isActive(): boolean
  isConnected(): boolean
  updateWebSocketUrl(newUrl: string): void
}

ConversationState

interface ConversationState {
  isConnected: boolean;
  isRecording: boolean;
  isPlaying: boolean;
  currentAmplitude: number;
  error: string | null;
}

AudioInterface

Low-level audio interface for custom implementations.

class AudioInterface {
  constructor(
    inputCallback: (audioData: ArrayBuffer | ArrayBufferView) => void,
    amplitudeCallback?: (amplitude: number) => void,
    onInterrupt?: () => void
  )

  async start(): Promise<void>
  stop(): void
  play(audioData: ArrayBuffer): void
  interruptPlayback(): void
  setAudioPlaybackState(isPlaying: boolean): void
  cleanup(): void
}

WebSocketManager

WebSocket communication manager.

class WebSocketManager {
  constructor(websocketUrl: string, audioInterface?: AudioInterface)

  connect(): Promise<void>
  disconnect(): void
  sendAudioData(audioData: ArrayBuffer): void
  sendMessage(message: object): void
  sendInterrupt(): void
  isConnected(): boolean
}

Examples

Basic Usage with State Monitoring

import { ConversationManager, ConversationState } from '@austack/client';

const conversation = new ConversationManager(
  'ws://localhost:8000/ws',
  (state: ConversationState) => {
    if (state.error) {
      console.error('Conversation error:', state.error);
    }
    
    if (state.isConnected) {
      console.log('Connected to server');
    }
    
    if (state.isRecording) {
      console.log('Recording audio...');
    }
    
    if (state.currentAmplitude > 0.1) {
      console.log('Voice detected');
    }
  }
);

try {
  await conversation.startConversation();
  console.log('Conversation started successfully');
} catch (error) {
  console.error('Failed to start conversation:', error);
}

Custom Audio Configuration

import { AudioInterface, WebSocketManager } from '@austack/client';

// Create custom audio interface
const audioInterface = new AudioInterface(
  (audioData) => {
    // Handle audio input
    webSocketManager.sendAudioData(audioData);
  },
  (amplitude) => {
    // Monitor voice amplitude
    console.log('Amplitude:', amplitude);
  },
  () => {
    // Handle interruptions
    console.log('User interrupted playback');
  }
);

const webSocketManager = new WebSocketManager('ws://localhost:8000/ws', audioInterface);

await webSocketManager.connect();
await audioInterface.start();

Audio Configuration

The SDK uses the following default audio settings:

• Sample Rate: 16kHz (input and output)
• Channels: Mono (1 channel)
• Format: 16-bit PCM
• Chunk Size: 1024 samples
• Silence Threshold: 0.01
• Silence Timeout: 2 seconds
• Send Interval: 0.5 seconds

These settings are optimized for speech recognition and conversation applications.

Browser Compatibility

This SDK is designed for modern browsers with support for:

• WebSocket API
• Web Audio API
• MediaDevices API (getUserMedia)

License

MIT