Setup: Twilio Elastic SIP Trunk → sip.retellai.com, two +57 (Colombia) numbers.
Problem (started 24-Jul, worked fine 23-Jul, no changes on our side): inbound phone calls
connect at the SIP layer and the agent’s greeting TTS is generated, but there is NO audio in
either direction and ZERO ASR events (user turns = 0). The caller hears nothing and hangs up
after a few seconds; the call is logged as user_hangup.
Why this looks like a LiveKit SBC / RTP issue, not our setup:
- Web/WebRTC test calls work perfectly today (132s, 22 bidirectional turns) → agent, LLM,
TTS and our backend are all fine. - Both +57 numbers fail identically (same trunk) → account/platform level.
- Twilio shows all calls completed, no errors, healthy balance → their side is fine.
- Matches known threads: “one-way audio”, “RTP not flowing outbound from LiveKit”,
“calls dropped at the LiveKit level (logged as user hangup)”.
Muted call IDs (today, 24-Jul):
- call_5c50a071ea095a3b374f8c60e81
- call_6e2903ac52fc822b5ed72448a3c
- call_3cb7f8ae29ac9964e5b6d15d0c6
Working call WITH audio (23-Jul): call_fcca1eb5f9efd6cd9ddb1e6679b
Questions:
- Is this the known one-way-audio / RTP-not-flowing-from-LiveKit regression?
- Is the LiveKit region serving our +57 numbers degraded? Can we be routed to another region?
- Workaround or ETA? This is production-blocking for a live business.
Agents: agent_2ea44a0ac6bbc83558e689107b, agent_0b95eea5ea977faa367294adea
Account: brainaia25@gmail.com
