Skip to main content

Linux Audio Plugin

Linux-specific audio implementation using naudiodon2 for recording, playback, transcription, and text-to-speech.

Overview​

The @tokenring-ai/linux-audio package provides a Linux-specific implementation of the AudioProvider interface using the naudiodon2 library for native audio operations. It enables recording, playback, transcription, and text-to-speech capabilities on Linux systems within the Token Ring AI framework.

Key Features​

  • Recording: Capture audio from microphone using naudiodon2
  • Playback: Play WAV audio files through system audio
  • Transcription: Convert audio to text using OpenAI Whisper
  • Text-to-Speech: Generate speech from text using OpenAI TTS
  • Format Support: WAV format for recording/playback, MP3 for TTS
  • Native Performance: Uses naudiodon2 for efficient audio I/O

Core Components​

LinuxAudioProvider​

Concrete implementation of AudioProvider for Linux systems.

Key Methods:

  • record(abortSignal, options?): Record audio from microphone
    • Returns path to recorded WAV file
    • Supports abort signal for stopping recording
  • transcribe(audioFile, options?): Convert audio to text using OpenAI Whisper
    • Supports various audio formats
    • Returns transcription text
  • speak(text, options?): Convert text to speech using OpenAI TTS
    • Generates MP3 audio file
    • Returns path to audio file
  • playback(filename, options?): Play audio files through system audio
    • Supports WAV format playback
    • Uses naudiodon2 for audio output

Installation​

Requires naudiodon2 native dependencies for Linux audio support:

# Install system dependencies (Ubuntu/Debian)
sudo apt-get install libasound2-dev

# Package is installed as part of Token Ring monorepo

Usage Example​

import { LinuxAudioProvider } from '@tokenring-ai/linux-audio';
import { AudioService } from '@tokenring-ai/audio';

// Create and register provider
const provider = new LinuxAudioProvider({
  sampleRate: 48000,
  channels: 1,
  format: 'wav'
});

const audioService = new AudioService();
audioService.registerProvider('linux', provider);
audioService.setActiveProvider('linux');

// Record audio
const abortController = new AbortController();
const recording = await audioService.record(abortController.signal);

// Transcribe audio
const transcription = await audioService.transcribe(recording);
console.log('Transcription:', transcription);

// Text-to-speech
const audioFile = await audioService.speak('Hello, world!');

// Playback
await audioService.playback(audioFile);

Configuration Options​

  • sampleRate: Audio sample rate in Hz (default: 48000)
  • channels: Number of audio channels (default: 1 for mono)
  • format: Audio format (default: 'wav')
  • OpenAI API key required for transcription and TTS (via environment variable)

Dependencies​

  • @tokenring-ai/audio@0.1.0: Abstract audio framework
  • @tokenring-ai/agent@0.1.0: Agent integration
  • naudiodon2@^3.2.0: Native audio I/O for Node.js
  • wav@^1.0.2: WAV file format support
  • @openai/openai@^4.0.0: OpenAI SDK for transcription and TTS

System Requirements​

  • Linux operating system
  • ALSA audio system (libasound2-dev)
  • OpenAI API key for transcription and TTS features