homeguild_voice_deepgram

Deepgram STT provider for homeguild_voice_kit. Real-time speech-to-text via WebSocket with multi-language support, Voice Activity Detection, and keep-alive.

Part of Voice Kit by HomeGuild Labs.

Usage

import 'package:homeguild_voice_deepgram/homeguild_voice_deepgram.dart';

final stt = DeepgramSTTProvider(
  apiKey: 'YOUR_DEEPGRAM_KEY',
  model: 'nova-2',       // Deepgram's latest model
  language: 'multi',      // Auto-detect English/Spanish
  sampleRate: 16000,
  onUsageTracked: (seconds) => print('Used ${seconds}s of STT'),
);

Pass it to VoiceConversationManager from homeguild_voice_kit:

final manager = VoiceConversationManager(
  sttProvider: stt,
  ttsProvider: tts,
  streamAdapter: myBackend,
);

Features

• Real-time transcription with interim and final results
• Multi-language auto-detection (English/Spanish via language: 'multi')
• Voice Activity Detection (VAD) events for natural pause detection
• Configurable model, language, encoding, and sample rate
• Keep-alive mechanism (5s intervals) for stable WebSocket connections
• Optional usage tracking callback

Requirements

• A Deepgram API key
• Microphone permission on the target platform

License

MIT — see LICENSE.