speech_deepgram

Deepgram streaming speech-to-text for the provider-neutral speech_core contracts.

The adapter accepts caller-owned PCM; it does not capture a microphone or own playback. Audio writes are bounded, credentials come from a renewable token source, and transport failures use stable SpeechFailure codes.

import 'package:speech_deepgram/speech_deepgram.dart';

final tokenSource = StaticDeepgramTokenSource(
  DeepgramAccessToken(value: apiKey),
);

Compose the provider with audio_flutter and audio_kit_graph to share one capture with recording, metering, VAD, or additional speech providers.

This release uses a dart:io WebSocket transport and does not support web.