Add AudioWorklet-based PCM16 streaming to VoiceRecorder with WebSocket
connection for near-real-time transcription. Segments appear as finalized
text that never shifts. Add speaker diarization button on audio notes
with color-coded speaker labels. Graceful fallback to batch transcription
when WebSocket unavailable.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add AUDIO to NoteType enum, duration field to Note model
- New VoiceRecorder component (MediaRecorder API, upload, transcribe)
- New /api/voice/transcribe proxy route to voice-command-api container
- Audio MIME types added to upload whitelist
- Audio player + transcript display on note detail page
- AUDIO type button on new note page with recorder UI
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Jeff Emmett