Hi! I’m experiencing a major issue with multiple agents on LiveKit Cloud. About 12 hours ago, all my agents stopped producing TTS audio during sessions. LLM text responses still appear in the transcript, but agents no longer publish any audio tracks. Affected agents: - phone_agent - onboarding_agent - Sandbox agents - Production agents Symptoms: - Session transcript shows correct assistant replies - But no TTS audio is published in the room - Session timeline shows “handoff to default_agent”, even though I don’t have a default_agent - Happens across multiple agents and environments - No config changes were made on my side Example broken session_id: RM_J9dyWVbmZHYS Everything worked fine yesterday — now agents are silent everywhere. This looks like a TTS/dispatch pipeline regression on LiveKit Cloud. Could someone please check the TTS subsystem, dispatch rules resolver, and agent audio publishing on the cluster?

I see the agent for that room came from your onboarding agent, so I think the

default_agent

might be a red herring. What TTS do you use? Is there perhaps an issue with credits? I don't see any other reports on our end.

we was using Cartesia. solved by changing TTS to ElevenLabs. errors that i was getting:

tts_node: APIError: no audio frames were pushed for text: Hi! I'm Alex ... I'm here to help! (body=None, retryable=True)
tts_request: APIError: no audio frames were pushed for text: Hi! I'm Alex from ... I'm here to help! (body=None, retryable=True)

🙌 I recommend taking a look at https://docs.livekit.io/agents/build/events/#fallbackadapter, to avoid downtime in the future

Ohh, thank you! This is exactly what I'm looking for🤝