WordloopML — Speaker Diarization
speaker_label populated on TranscriptSegmentProducedEvent from AssemblyAI diarization, and SegmentFeaturesProducedEvent for voice embeddings.
ML — Speaker Diarization
Owner: ML Engineer Domain: ml Complexity: S Prerequisite: Milestone 04 merged
When this slice is complete, every TranscriptSegmentProducedEvent ML emits carries the speaker_label AssemblyAI's diarization assigns, and ML emits SegmentFeaturesProducedEvent carrying voice embeddings so a future session can suggest identity across meetings.
Required Capabilities
-
TranscriptSegmentProducedEventcarriesspeaker_labelpopulated from AssemblyAI's diarization output. - ML emits
SegmentFeaturesProducedEventwith a voice embedding for each segment.
Dependencies
- Milestone 04 (
live-transcript-to-final-rebuild) must be merged — the streaming AssemblyAI proxy andTranscriptSegmentProducedEventcontract.
Domain Notes
Coverage gap. No bet-progress test exercises ML's real diarization or voice-feature output. test_milestone_09_speaker_labelling.py's module docstring states the suite instead pushes segments "the way the ML service does it" — POST /transcriptions/{id}/segments with the service token, using hand-written A/B labels — rather than letting a real (or even mocked-diarizing) AssemblyAI session assign them. This proves Core's label→person mapping and its survival through replacement thoroughly, but the actual diarization step (does AssemblyAI's speaker_label output get faithfully carried onto TranscriptSegmentProducedEvent?) is unverified by the bet suite. SegmentFeaturesProducedEvent / voice embeddings are not referenced anywhere in the bet suite at all — permanent ML unit tests (services/wordloop-ml/tests/unit/core/services/test_voice.py, test_speaker_identification.py) may cover this in isolation, but there is no bet-progress acceptance test for it.
Test Cases
Not yet covered by a bet-progress test. There is no test in tests/bets/meeting-recording/test_milestone_09_speaker_labelling.py (or elsewhere in the bet suite) that exercises real AssemblyAI diarization output reaching TranscriptSegmentProducedEvent, or SegmentFeaturesProducedEvent at all.
| Test | Location | Assertion |
|---|---|---|
| (none — not yet covered) | — | TranscriptSegmentProducedEvent.speaker_label reflecting AssemblyAI's own diarization (as opposed to a test posting hand-written labels directly) is not exercised by any bet-progress test. |
| (none — not yet covered) | — | SegmentFeaturesProducedEvent / voice embeddings are not exercised by any bet-progress test. |
Completion Checklist
- Code merged and deployed
- Bet progress tests pass (
./dev test bet meeting-recording) - Permanent service tests implemented per testing strategy
- Code review completed
- Testing review completed
- System documentation updated (as applicable)
Core — Speaker Assignment
POST /meetings/:id/speaker-labels, person lookup/create, speaker_label to person_id mapping persisted to transcript_segments and surviving PUT /transcriptions/:id/segments replacement, and SpeakerStateUpdatedEvent to ML.
10 Audio Playback And Review
Final review supports audio playback, signed access, and synced transcript navigation.