I did a session by myself and I am looking at the Agent Analytics page. Do you think performance will be affected if the Anam Agent joins in US Central and the LiveKit server is in US East?

Thanks for reaching out, Ganesh. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

From the looks of what I researched, it said the following: Deepgram utilizes cloud infrastructure, with specific documentation pointing to resources provisioned in the GCP us-west1 region. ElevenLabs’ default infrastructure for the Americas routes to US-Central for processing. Probably the issue for a lot of the latency I am experiencing then huh?

You’re on the right track — those regional differences likely explain much of the latency you’re seeing. Why: • Each hop to a differently located service adds network and processing latency. Deepgram in GCP us-west1 and ElevenLabs routing to US‑Central means your audio is traveling between regions before/after the agent processes it. • Agent⇄model (STT/LLM/TTS) latency typically contributes far more to turn time than client⇄LiveKit server latency, so mismatched model locations will show up as the largest source of delay. What to try next: • Align regions where possible: host the agent and any models (or prefer model endpoints) in the same region as each other to cut internal API round trips. • Run measured tests: capture per-stage latency (STT, LLM, TTS, network) from Agent Observability or per‑turn metrics so you can quantify which provider/step adds the most time. • Short-term mitigation: if you can’t move services, try routing the LiveKit server to a region that reduces total round trips (choose a compromise region) or use lower-latency model endpoints if available. If you want, share a sample turn’s latency breakdown (STT, LLM, TTS, and network times) and I’ll help interpret where to focus optimizations. , with interactive elements