fix: commit user turn with STT and realtime #4663

chenghao-mou · 2026-01-30T17:38:34Z

Turns out using STT with a realtime model will trigger a response when committing a turn. I didn't notice the interruption when testing so wrongly assumed it was doing just fine.

Example to reproduce:

    @session.on("agent_state_changed")
    def _on_agent_state_changed(ev: AgentStateChangedEvent):
        if ev.new_state == "speaking":
            logger.info("speaking")
            session.input.set_audio_enabled(False)
        elif ev.new_state == "listening":
            logger.info("listening")
            session.input.set_audio_enabled(True)
            session.clear_user_turn()

            async def commit_user_turn():
                nonlocal session
                logger.info("waiting 20 seconds to commit user turn")
                await asyncio.sleep(20)
                logger.info("committing user turn")
                session.commit_user_turn()
                logger.info("user turn committed")
                session.input.set_audio_enabled(False)

            task = asyncio.create_task(commit_user_turn())
            tasks.add(task)
            task.add_done_callback(tasks.discard)
        else:
            logger.info(f"agent state changed to {ev.new_state}")

The model will respond after 20 seconds

the STT transcripts will be committed into the chat context without triggering a duplicate response
the model will only respond once when committing a turn manually

cc @bml1g12

Summary by CodeRabbit

Breaking Changes
- Removed the commit_user_turn method from the Realtime Session interface. This method was previously unused and non-functional across implementations and has been eliminated to simplify the API surface and reduce unnecessary complexity. Applications that reference this method require updating for compatibility.

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

coderabbitai · 2026-01-30T17:38:56Z

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

🔍 Trigger a full review

📝 Walkthrough

Walkthrough

This PR removes the commit_user_turn() abstract method from the RealtimeSession interface and eliminates all corresponding implementations and call sites across the livekit-agents and livekit-plugins repositories, including AWS, Google, OpenAI, and Ultravox integrations.

Changes

Cohort / File(s)	Summary
RealtimeSession Interface `livekit-agents/livekit/agents/llm/realtime.py`	Removed abstract method `commit_user_turn(self) -> None` from the base RealtimeSession interface.
Realtime Agent Control `livekit-agents/livekit/agents/voice/agent_activity.py`	Removed invocation of `self._rt_session.commit_user_turn()` in the `commit_user_turn` method; now only delegates to AudioRecognition.
AWS Plugin Implementation `livekit-plugins/livekit-plugins-aws/livekit/plugins/aws/experimental/realtime/realtime_model.py`	Removed `commit_user_turn()` method that logged a not-supported warning for Nova Sonic's Realtime API.
Google Plugin Implementation `livekit-plugins/livekit-plugins-google/livekit/plugins/google/realtime/realtime_api.py`	Removed `commit_user_turn()` method that logged a not-supported warning for Gemini Realtime API.
OpenAI Plugin Implementations `livekit-plugins/livekit-plugins-openai/livekit/plugins/openai/realtime/realtime_model.py`, `livekit-plugins/livekit-plugins-openai/livekit/plugins/openai/realtime/realtime_model_beta.py`	Removed `commit_user_turn()` methods that checked auto-response settings, logged warnings, called `commit_audio()`, and emitted ResponseCreateEvent with empty parameters.
Ultravox Plugin Implementation `livekit-plugins/livekit-plugins-ultravox/livekit/plugins/ultravox/realtime/realtime_model.py`	Removed `commit_user_turn()` method that logged a not-supported warning.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~22 minutes

Possibly related PRs

AGT-2474: add commit user turn support for realtime models #4622: Directly conflicts with this PR; that PR adds commit_user_turn to RealtimeSession while this PR removes it, representing an inverse change.

Suggested reviewers

bml1g12
longcw

Poem

🐰 A method exits stage left with grace,
No more turns to commit to this place,
From interfaces clean to plugins refined,
Simpler flows left behind,
One less hop in the agent's embrace! 🚀

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Title check	⚠️ Warning	The PR title mentions 'commit user turn with STT and realtime' but the actual changes remove commit_user_turn methods across all realtime implementations and call sites, not add or fix functionality with STT. The title is misleading about the nature of the changes.	Consider updating the title to accurately reflect the changes, such as 'refactor: remove commit_user_turn from realtime session implementations' or 'chore: remove unused commit_user_turn method from realtime models'.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

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

✨ Finishing touches

🧪 Generate unit tests (beta)

Create PR with unit tests
Post copyable unit tests in a comment
Commit unit tests in branch fix/remove-commit-user-turn

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

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

davidzhao

is this true when manual turn mode is used?

I see that if it's not server vad, we have to manually create them: https://platform.openai.com/docs/api-reference/realtime-client-events/input_audio_buffer/commit

chenghao-mou · 2026-01-30T17:59:27Z

is this true when manual turn mode is used?

I see that if it's not server bad, we have to manually create them: platform.openai.com/docs/api-reference/realtime-client-events/input_audio_buffer/commit

I think I found the issue: when using with STT, it actually committed the transcripts to the model and therefore triggered a response. Without STT, it will not respond as expected.

davidzhao · 2026-01-30T18:18:09Z

is this true when manual turn mode is used?
I see that if it's not server bad, we have to manually create them: platform.openai.com/docs/api-reference/realtime-client-events/input_audio_buffer/commit

I think I found the issue: when using with STT, it actually committed the transcripts to the model and therefore triggered a response. Without STT, it will not respond as expected.

got it, so the bug is: when STT is used with manual turn detection mode and realtime model, it should not trigger a response.

chenghao-mou · 2026-01-30T18:30:52Z

is this true when manual turn mode is used?
I see that if it's not server bad, we have to manually create them: platform.openai.com/docs/api-reference/realtime-client-events/input_audio_buffer/commit

I think I found the issue: when using with STT, it actually committed the transcripts to the model and therefore triggered a response. Without STT, it will not respond as expected.

got it, so the bug is: when STT is used with manual turn detection mode and realtime model, it should not trigger a response.

Yep. But it is also kinda ambiguous what users want if they have an STT configured: do they want it only for the transcripts or actual model text input.

bml1g12 · 2026-01-30T20:18:52Z

In our use case we want STT with manual turn taking both as a way to know when it's safe to commit audio before a user turn has ended (as a user turn could be very long, we cannot wait until end of turn to commit) and for its improved transcripts post call versus openai server side transcription

So I think ideally we'd have STT transcripts available and in the local chat context (so if there is a connection error we can restore the conversation on a fresh session) but the remote openai realtime chat context need not even be aware of it as it can function perfectly fine without a transcript of user utterances. STT also provides a neat signal to the user that the system is working during a long user turn (by allowing them to view their own transcription as they talk).

devin-ai-integration

✅ Devin Review: No Issues Found

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

View in Devin Review to see 5 additional flags.

chenghao-mou requested a review from a team January 30, 2026 17:38

chenghao-mou changed the title ~~remove commit user turn for realtime models~~ Revert: remove commit user turn for realtime models Jan 30, 2026

chenghao-mou mentioned this pull request Jan 30, 2026

AGT-2474: add commit user turn support for realtime models #4622

Merged

davidzhao reviewed Jan 30, 2026

View reviewed changes

chenghao-mou changed the title ~~Revert: remove commit user turn for realtime models~~ fix: commit user turn with STT and realtime Jan 30, 2026

chenghao-mou marked this pull request as draft January 30, 2026 19:13

chenghao-mou marked this pull request as ready for review February 1, 2026 11:03

chenghao-mou requested a review from a team February 1, 2026 11:03

devin-ai-integration bot reviewed Feb 1, 2026

View reviewed changes

chenghao-mou added 4 commits February 1, 2026 11:07

remove commit user turn for realtime models

4840496

temp

1954c49

allow stt transcript flush without triggering a response

d579a4f

fix early return

a3a206e

chenghao-mou force-pushed the fix/remove-commit-user-turn branch from 6099024 to a3a206e Compare February 1, 2026 11:07

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix: commit user turn with STT and realtime #4663

fix: commit user turn with STT and realtime #4663

Uh oh!

chenghao-mou commented Jan 30, 2026 •

edited

Loading

Uh oh!

coderabbitai bot commented Jan 30, 2026 •

edited

Loading

Review skipped

Walkthrough

Changes

Estimated code review effort

Possibly related PRs

Suggested reviewers

Poem

Uh oh!

davidzhao left a comment

Uh oh!

chenghao-mou commented Jan 30, 2026

Uh oh!

davidzhao commented Jan 30, 2026

Uh oh!

chenghao-mou commented Jan 30, 2026

Uh oh!

bml1g12 commented Jan 30, 2026 •

edited

Loading

Uh oh!

devin-ai-integration bot left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

fix: commit user turn with STT and realtime #4663

Are you sure you want to change the base?

fix: commit user turn with STT and realtime #4663

Uh oh!

Conversation

chenghao-mou commented Jan 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary by CodeRabbit

Uh oh!

coderabbitai bot commented Jan 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Walkthrough

Changes

Estimated code review effort

Possibly related PRs

Suggested reviewers

Poem

Uh oh!

davidzhao left a comment

Choose a reason for hiding this comment

Uh oh!

chenghao-mou commented Jan 30, 2026

Uh oh!

davidzhao commented Jan 30, 2026

Uh oh!

chenghao-mou commented Jan 30, 2026

Uh oh!

bml1g12 commented Jan 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

devin-ai-integration bot left a comment

Choose a reason for hiding this comment

✅ Devin Review: No Issues Found

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

chenghao-mou commented Jan 30, 2026 •

edited

Loading

coderabbitai bot commented Jan 30, 2026 •

edited

Loading

bml1g12 commented Jan 30, 2026 •

edited

Loading