Hi all, I would like to reach out and see if anyone has encountered similar issue of an agent suddenly (and randomly) stop responding - We have a phone integrated pipeline agent that in some sessions would suddenly stop responding to user. From the user's perspective, they can't hear anything from the agent. However, the agent log shows it is still responding, in the sense that it's getting transcript and generating audio through LLM and TTS as usual. The egress recording has the agent audios recorded as well. During these "non responding" period, all the livekit room logs in the form of "[LiveKitRoomPoolManager] : Forwarded text stream message to connection xxx" logs that keeps showing up during normal functioning are no longer being logged. Thanks a lot!

What version of LiveKit are your using?

pip freeze |grep livekit

When did you first notice this to start happening? Are there steps i can perform to reproduce this using basic_agent.py? Can you reproduce the issue using basic_agent?

Thank you so much for the reply @refined-appointment-81829! We're using livekit cloud. Not very sure about the version. First happened on 14th Aug. Then two more times on 27th Aug. I can't reproduce them unfortunately. What I can see consistently is that logs like below are available during normal phone call, and during the agent non-response period, there's also none of these logs.

[info]: [LiveKitRoomPoolManager] Received transcription from agent xxxx
[debug]: [LiveKitRoomPoolManager] Forwarded text stream message to connection xxxx
[info] [LiveKitRoomPoolManager] Received transcription from pool-worker-xxxx
[debug]: [LiveKitRoomPoolManager] Forwarded text stream message to connection xxxx

@refined-appointment-81829 Is their any cases, where Agent audio is transmitted to livekit cloud properly. But user not able to listen agent audio either due to twilio issue or livekit cloud to backend server transfer issue through to webrtc ? if yes, is their any mechanism to identify it through some events ?