original idea #2476 / thank @0xrushi

my comments:

---
3/ the flow should be: users setting up their own STT service (name, url, etc) via the omi app -> either the app sends these params along with the /listen API request or the /listen API checks the current active STT service configured by users -> forward the audio bytes to the STT service. right?
4/ if yes, it would be better if

• you could keep the omi backend super clean. new custom_stt_service just like deepgram or soniox, an utils/custom_stt
• your repo implements the custom stt service interfaces (apis)
• add support for the ability to config the active custom stt to the app and the backend