fix(openai): continue GPT Live only once every backend call is answered

[longcw]

Problem: GPT Live refuses response.create while any function call in the session is unanswered, but the plugin gated each continuation on one response per delegation. A second response.created on the same delegation replaced that record, so an earlier call stopped counting, the early continuation was refused, and the backend never answered again.

Fix: The gate is session-wide: a delegation holds only the calls of the response still running, and confirmed calls wait in one session-wide set. One response.create goes out when nothing runs, nothing is open and an answer went out since the last continuation, and a failed response drops its own calls.

Ports livekit/agents#7266.

Context for reviewing and coding agents

How to see it

holds the continuation until every open call in the conversation is answered runs two responses under one delegation and answers the second call first. On main the plugin sends response.create at that point, so the socket receives three events where the test expects two. The second new test reaches the same state through response.failed.

Why the running response is still tracked per delegation

response.output_item.done carries no response id, so a call is attributed to the newest response.created under its delegation. That attribution is what lets a failed response drop exactly its own calls. The map costs nothing while it holds one entry.

What this port leaves out

The Python fix also warns when a tool result carries reply_required=False, because GPT Live speaks the result anyway. llm.FunctionCallOutput in this repo has no replyRequired field: it lives on _SanitizedOutput in agents/src/voice/generation.ts:253 and never reaches _appendItems, and adding it would change the public API.

Divergence from the Python diff

livekit/agents#7266 moves the call id and function name out of the PII log fields. This repo checks for untagged PII logs, so the lk.pii. tags stay as they are. livekit/agents#7230, the duplicate and incomplete call guard that the Python diff shows as context, is not ported here either.

Blast radius

Only GPTLiveSession in plugins/openai/src/realtime/gpt_live_model.ts changes, and the removed DelegatedResponse interface was private, so pnpm api:check for the plugin reports no API change. The other 80 tests in the file pass without edits.

[changeset-bot]

🦋 Changeset detected

Latest commit: b7b3be7

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

This PR includes changesets to release 39 packages

Name	Type
@livekit/agents-plugin-openai	Patch
@livekit/agents-plugin-anam	Patch
@livekit/agents-plugin-cartesia	Patch
@livekit/agents-plugin-cerebras	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-neuphonic	Patch
@livekit/agents-plugin-perplexity	Patch
@livekit/agents-plugin-rime	Patch
@livekit/agents-plugin-sarvam	Patch
@livekit/agents-plugin-xai	Patch
@livekit/agents	Patch
@livekit/agents-plugin-anthropic	Patch
@livekit/agents-plugin-assemblyai	Patch
@livekit/agents-plugin-azure	Patch
@livekit/agents-plugin-baseten	Patch
@livekit/agents-plugin-bey	Patch
@livekit/agents-plugin-deepgram	Patch
@livekit/agents-plugin-did	Patch
@livekit/agents-plugin-lemonslice	Patch
@livekit/agents-plugin-liveavatar	Patch
@livekit/agents-plugin-livekit	Patch
@livekit/agents-plugin-meta	Patch
@livekit/agents-plugin-minimax	Patch
@livekit/agents-plugin-mistral	Patch
@livekit/agents-plugin-mistralai	Patch
@livekit/agents-plugin-phonic	Patch
@livekit/agents-plugin-protoface	Patch
@livekit/agents-plugin-resemble	Patch
@livekit/agents-plugin-runway	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

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.