Great question — answering as the UphillTech team:
Im working on something similar and even when I couldn't test this myself because of the time (it takes to replicate it), I'm just comparing what you did versus what I did Looking at your code, I suspect the main issue is with the track parameter . When you're sending audio back to the caller, you're using: "track": "inbound" # ❌ This is incorrect for sending audio TO the callerT "track": "outbound" # ✅ Use this when sending audio TO the caller Remember: "inbound" = audio FROM the caller TO your app "outbound" = audio FROM your app TO the caller Since you want the caller to hear the AI response, you need "outbound" .
Full answer by lapiceroazul4: https://stackoverflow.stackexchange.com/a/79700104 (CC BY-SA 4.0).
Need it done for your business? UphillTech.
