i've tried both models but when making a call the models dont respond to voice - Ican see the users messages in the call transcript logs. Switching models to other LLMs makes it work again

Hey! To help track down this issue, could you share: - The call ID This would really help us figure out what went wrong!

sure - 6309efff-79e3-46a8-afa2-c2fff3d35a27

Hey @yandos666 sorry for the delay, it is related to endpoting which controls how long to wait for user to finish speaking by default is uses a value of 10 seconds which isn't fruitful so it's expected to use 300ms instead. Here's what you need to make change to your transcriber:

{
  "model": "nova-2",
  "language": "en",
  "provider": "deepgram",
  "endpointing": 300
}