Repository navigation
fix(livekit): show end-of-turn timing - #1132
Merged
Merged
Conversation
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
| and delay >= 0 | ||
| ): | ||
| wait = _new( | ||
| user, "livekit.eou_wait", start_time=stopped, metadata={"contrib.livekit.outcome": "committed"} |
There was a problem hiding this comment.
this feels like a more generic span we should spec out, eou_wait will apply to basically every other framework out there as well. But I'm fine with this shipping in current state to unblock as well.
Contributor
Author
There was a problem hiding this comment.
I think that's a fine idea. We can migrate it to our schema later.
Abhijeet Prasad (AbhiPrasad)
approved these changes
Nov 17, 2026
David Elner (delner)
force-pushed
the
delner/sdk-851-2-livekit-model-spans
branch
from
November 18, 2026 04:14
01ddde0 to
601ba26
Compare
David Elner (delner)
force-pushed
the
delner/sdk-851-3-livekit-latency
branch
from
November 18, 2026 04:14
8f55958 to
e68687c
Compare
Base automatically changed from
delner/sdk-851-2-livekit-model-spans
to
main
November 18, 2026 04:23
> 🤖 **This text was AI-generated.** Voice traces lacked the native timing needed to explain pauses between a caller's speech and the agent's response. Accepted user turns now include available end-of-turn and transcription delays. A livekit.eou_wait task shows the observed interval from speech end to the accepted decision. ```text user_turn user_speech -> native VAD silence/inference diagnostics livekit.eou_wait -> observed endpointing interval ``` The native user-turn callback duration is attributed through the existing turn context. Fields use contrib.livekit metadata; missing or invalid values are omitted. Model timing and usage remain on model children. Refs: https://linear.app/braintrustdata/issue/SDK-832
David Elner (delner)
force-pushed
the
delner/sdk-851-3-livekit-latency
branch
from
November 18, 2026 04:24
e68687c to
4877c0e
Compare
David Elner (delner)
enabled auto-merge (squash)
November 18, 2026 04:29
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
Voice traces do not clearly explain the wait between a user finishing speaking and LiveKit deciding that their turn is complete.
Solution
Add LiveKit’s turn-detection and transcription timings to the relevant turn. Show the wait after speech ends as a visible step:
This makes it easier to understand where response time is spent.