Call transcript and audio not available

I have an agent running in production deployed on livekit cloud using nodejs SDK.
Few calls were reported poor by users and I have debug those call to see exactly what happend but I can’t see audio recordign and transcript of those calls on the livekit cloud.
What could be the reason for this? Is this some bug in livekit cloud or I am missing something. majority calls have this but very few calls do not have audio and transcript of the call.

@Munam_Tariq,

Worth separating: transcript comes from the agent session (your worker captures it). Audio recording is egress (separate subsystem, often auto-egress on Cloud). Both missing on the same call usually means something dropped before either pipeline ran.

Most likely patterns:

  • Agent didn’t run on that call: worker failed to dispatch, crashed early, or didn’t pick up the job. Worker logs for that room ID will show it.

  • Caller disconnected before audio flowed: mic denied, instant hangup, or network failure. Sessions row shows ~0s of media.

  • Auto-egress transient failure: occasional capacity or config errors. Check the Egress tab for failed entries on those rooms and egress_ended webhooks if you wire them.

Grab one specific room ID from a missing call and cross-check the Sessions dashboard, Egress tab, and worker logs. That triangulation almost always names which subsystem dropped it.

call duration is around 7 minutes and call happend, user talked to agent. I know it because it was internal user.

@Munam_Tariq, OK, that rules out the first two from my #2. Two remaining candidates given “majority works, a few don’t”:

Observability rate-limit drop: Cloud caps agent observability ingestion at 1,000 events/min. Bursty or concurrent traffic can push individual calls past the per-project window, and dropped events leave the dashboard showing a session row without the linked transcript and audio attachments. Worker logs in that case would show 429 QuotaStatusNetworkError against production.livekit.cloud/observability/logs/otlp/v0.

Auto-egress finalize failure: recording started but failed at upload or segment-merge. Check the Egress tab for that room’s egress ID and terminal status, and the egress_ended webhook payload if wired.

To unblock this fast: grab one specific room ID and ping @CWilson or @darryncampbell. They have Cloud-side log access that can name the dropped subsystem on that exact room within minutes.

What is the Room Sesion id that is missing the recording?

I assume you have observablity enabled?