fix: agent state using QueueAudioOutput / DataStreamAudioOutput

[CathyL0]

Summary by CodeRabbit

• Bug Fixes

  • Improved reliability of asynchronous voice processing with safer cancellation and error handling.
  • More accurate detection of audio playback start so playback is signaled promptly on first frame.

• Improvements

  • Agent transitions to "thinking" earlier during voice interactions, giving faster user feedback.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

📝 Walkthrough

Walkthrough

Introduces an earlier "thinking" state transition during pipeline replies, adds defensive guards and narrower exception handling when reading futures, refactors agent state-change emission, and triggers playback_started callbacks on first audio frame for queue and datastream audio outputs.

Changes

Cohort / File(s)	Summary
Agent activity livekit/agents/voice/agent_activity.py	Move agent state to "thinking" earlier in the pipeline reply flow; add guards to async callbacks to check futures for cancellation/done before calling result(); narrow exception handling from BaseException to asyncio.CancelledError and Exception.
Agent session livekit/agents/voice/agent_session.py	Refactor emission of state-change events to build an AgentStateChangedEvent in a temporary variable before calling emit (semantically identical).
Audio playback — queue output livekit/agents/voice/avatar/_queue_io.py	Import time and trigger on_playback_started(created_at=time.time()) when the first audio frame is captured (QueueAudioOutput is end of chain).
Audio playback — datastream output livekit/agents/voice/avatar/_datastream_io.py	Import time and trigger on_playback_started(created_at=time.time()) when stream_writer is first established so first-frame futures are resolved proactively.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Suggested reviewers

• davidzhao
• longcw
• rimiakter5612-arch

Poem

🐰 I hopped in early to start the "think",
Futures now guarded before they blink,
Events wrapped tidy, playback sings true,
Small changes, big care — a rabbit's review! 🎵

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 45.45% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately identifies the main changes affecting agent state handling when using QueueAudioOutput and DataStreamAudioOutput components.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}