fix: block preemptive generation during tool execution

[Raysharr]

Description

Fix a race condition where preemptive generation can fire while tool execution is still in progress, causing LLM chat context corruption and session crashes.

Changes Made

• Add _toolExecutionInProgress flag to AgentActivity that is set true when performToolExecutions starts and false when tool execution completes (both normal and interrupt paths)
• Add guard check for _toolExecutionInProgress in onPreemptiveGeneration() to block preemptive generation while tools are executing
• Cover both reply task code paths (legacy and new)
• Add unit tests for the new guard behavior

Root Cause

When _markGenerationDone() fires after TTS playout (line 2107) but before await executeToolsTask.result (line 2108), mainTask clears _currentSpeech. This opens a window where onPreemptiveGeneration() passes all guards and snapshots chatCtx while it contains orphaned FunctionCall items (no matching FunctionCallOutput).

removeInvalidToolCalls() in utils.ts strips the orphans, producing "function call missing the corresponding function output, ignoring" warnings. The corrupted context then causes consecutive LLM generation failures.

Addresses TODO(AJS-272).

Pre-Review Checklist

• Build passes: All builds (lint, typecheck, tests) pass locally
• AI-generated code reviewed: Removed unnecessary comments and ensured code quality
• Changes explained: All changes are properly documented and justified above
• Scope appropriate: All changes relate to the PR title

Testing

• Automated tests added/updated
• All tests pass
• 2 new unit tests verify the _toolExecutionInProgress guard in onPreemptiveGeneration()
• Observed in production: session outbound-1774267632790 crashed with this exact sequence

Additional Notes

Timeline from production logs:

12:08:05 | available_slots API call started
12:08:06 | API completed, filler still playing
12:08:06 | [STT] patient speaks during tool execution
12:08:07 | preemptive generation starts ← should have been blocked
12:08:07 | ⚠️ function call missing the corresponding function output, ignoring
12:08:10–14 | ❌ failed to generate LLM completion (3x) → session crash

The fix is minimal and surgical — a single boolean flag checked in the existing guard clause. No changes to tool execution flow, scheduling, or _markGenerationDone() ordering.

[changeset-bot]

🦋 Changeset detected

Latest commit: 4bc2a97

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

This PR includes changesets to release 22 packages

Name	Type
@livekit/agents	Patch
@livekit/agents-plugin-anam	Patch
@livekit/agents-plugin-baseten	Patch
@livekit/agents-plugin-bey	Patch
@livekit/agents-plugin-cartesia	Patch
@livekit/agents-plugin-deepgram	Patch
@livekit/agents-plugin-elevenlabs	Patch
@livekit/agents-plugin-google	Patch
@livekit/agents-plugin-hedra	Patch
@livekit/agents-plugin-inworld	Patch
@livekit/agents-plugin-lemonslice	Patch
@livekit/agents-plugin-livekit	Patch
@livekit/agents-plugin-neuphonic	Patch
@livekit/agents-plugin-openai	Patch
@livekit/agents-plugin-phonic	Patch
@livekit/agents-plugin-resemble	Patch
@livekit/agents-plugin-rime	Patch
@livekit/agents-plugin-sarvam	Patch
@livekit/agents-plugin-silero	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