Livekit is taking time to respond to calls

Getting Started
1

Hi LiveKit team,

On LiveKit Cloud (livekit-agents 1.5.17), we have a consistent delay before the agent responds on every single call — both browser/WebRTC and SIP, even a single isolated call with no other load. So it’s structural and in the agent response path, not telephony- or load-related.It was working fine few hours back

Setup

  • LiveKit Cloud;
  • STT Deepgram, LLM OpenRouter/OpenAI, TTS ElevenLabs
  • turn_detection=“stt”; interruption detector logs use_proxy: true, inference_timeout: 0.7
2

@Vishalli_Maini I know you said this was every single call, but can you give me some call IDs of calls that experienced the issue, so I know I’m definitely looking at the right ones

3

SCL_VZtvEhB7278f This is the call id .

4

@Vishalli_Maini thanks, we have published an incident at LiveKit Status - Elevated latency in Hyderabad region, please subscribe to updates there.

5

@darryncampbell THANKS!

6

I also had some issues with latency. I ended up adding langfuse observation to see what was going on which in the end was that I had deployed that agent to the Indian DC which for whatever reason had some latency.

Through langfuse I also found other opportunities for reducing latency by better managing agent tool calls, the bonus also being the ability to improve call quality.

I would as part of any prod release advocate for some observability so you’re not stuck asking these types of questions and you know. Not an accusation BTW just a different way to be able to sleep at night knowing you have the ability to self-support.