Since there’s no ME region yet, it comes down to optimizing the path and response speed. You can reduce latency by routing calls through the closest available region (via Twilio/SIP), using a lightweight middleware server near your users to handle API calls, and tightening things like streaming, VAD, and response chunking so the assistant starts speaking faster instead of waiting on full outputs. I’ve implemented this kind of setup before and can help you get it running with minimal delay if you’d like. @raspberrytech