feat(assemblyai): surface end-of-turn confidence metadata

[rosetta-livekit-bot]

Summary

• Surface AssemblyAI streaming end-of-turn confidence on SpeechData.metadata for interim, preflight, and final transcripts.
• Preserve 0 as a meaningful confidence value while leaving metadata unset when the server omits the field.
• Add a patch changeset for @livekit/agents-plugin-assemblyai.

Port of livekit/agents#7107.

Source diff coverage

• Adapted: livekit-plugins/livekit-plugins-assemblyai/livekit/plugins/assemblyai/stt.py → plugins/assemblyai/src/stt.ts. Extended the existing target metadata mapper so end_of_turn_confidence is exposed as metadata.assemblyai.endOfTurnConfidence, following target provider namespacing and camelCase conventions. Existing event construction already applies this mapper to interim, preflight, and final transcripts.
• Adapted: tests/test_plugin_assemblyai_stt.py → plugins/assemblyai/src/stt-metadata.test.ts. Ported all four source cases through the target local WebSocket harness: interim confidence, final confidence, zero confidence, and absent confidence.
• Not applicable: None. Every source diff file has a target counterpart and was ported.

Testing

• pnpm exec vitest run plugins/assemblyai (10 passed, 1 credential-gated test skipped)
• pnpm exec turbo run build --filter=@livekit/agents-plugin-assemblyai
• pnpm --filter @livekit/agents-plugin-assemblyai lint
• pnpm exec prettier --check plugins/assemblyai/src/stt.ts plugins/assemblyai/src/stt-metadata.test.ts .changeset/surface-assemblyai-turn-confidence.md
• git diff --check

---

Ported from livekit/agents#7107

Original PR description

Summary

Surfaces the AssemblyAI streaming end_of_turn_confidence on SpeechData.metadata so callers can read it off the SpeechEvent (e.g. in an stt_node override) and threshold it for preemptive / eager LLM generation. The plugin already parsed the field from each Turn message but only logged it, so it never reached SpeechData. It's now attached to the interim, preflight, and final events, guarded on is not None (an early-partial 0.0 still surfaces; models that don't emit the field are unaffected).

Value semantics are model-defined:

• universal-3-5-pro / universal-3-6-pro / u3-rt-pro — rises from 0 toward 1 across the partials held between min_turn_silence and max_turn_silence, and is 1.0 on the final.
• universal-streaming-english / universal-streaming-multilingual — a per-partial end-of-turn probability on every partial, with the model's own score on the final.

Testing

• tests/test_plugin_assemblyai_stt.py: 4 new tests (interim, final, the 0.0-still-surfaces case, and the absent-field case). Full file: 115 passed; ruff and strict mypy clean.
• Verified against the production AssemblyAI streaming API on an ambiguous-ending clip: universal-3-5-pro and universal-3-6-pro surface the held-turn ramp (0.0 → 0.4 → 0.75 → 1.0), and universal-streaming-english / universal-streaming-multilingual surface their per-partial EOT probability — all readable off SpeechData.metadata.

[changeset-bot]

🦋 Changeset detected

Latest commit: 286683f

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 38 packages

Name	Type
@livekit/agents-plugin-assemblyai	Patch
@livekit/agents	Patch
@livekit/agents-plugin-anam	Patch
@livekit/agents-plugin-anthropic	Patch
@livekit/agents-plugin-azure	Patch
@livekit/agents-plugin-baseten	Patch
@livekit/agents-plugin-bey	Patch
@livekit/agents-plugin-cartesia	Patch
@livekit/agents-plugin-cerebras	Patch
@livekit/agents-plugin-deepgram	Patch
@livekit/agents-plugin-did	Patch
@livekit/agents-plugin-elevenlabs	Patch
@livekit/agents-plugin-fishaudio	Patch
@livekit/agents-plugin-google	Patch
@livekit/agents-plugin-hume	Patch
@livekit/agents-plugin-inworld	Patch
@livekit/agents-plugin-krisp	Patch
@livekit/agents-plugin-lemonslice	Patch
@livekit/agents-plugin-liveavatar	Patch
@livekit/agents-plugin-livekit	Patch
@livekit/agents-plugin-minimax	Patch
@livekit/agents-plugin-mistral	Patch
@livekit/agents-plugin-mistralai	Patch
@livekit/agents-plugin-neuphonic	Patch
@livekit/agents-plugin-openai	Patch
@livekit/agents-plugin-perplexity	Patch
@livekit/agents-plugin-phonic	Patch
@livekit/agents-plugin-protoface	Patch
@livekit/agents-plugin-resemble	Patch
@livekit/agents-plugin-rime	Patch
@livekit/agents-plugin-runway	Patch
@livekit/agents-plugin-sarvam	Patch
@livekit/agents-plugin-silero	Patch
@livekit/agents-plugin-soniox	Patch
@livekit/agents-plugin-tavus	Patch
@livekit/agents-plugins-test	Patch
@livekit/agents-plugin-trugen	Patch
@livekit/agents-plugin-xai	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.