Closed fakhirali closed 2 months ago
Now can identify when the interruption occurred. Need to send that back into the llm. Also the say_multiple_stream function need to be refactored heavily. Contains lots of repeated code.
The llm now gets '...' when an interruption happens. The ellipses get added after the phrase that was completely said. For future work we could add it after the exact word where the interruption occurred.
Right now just sends the interrupt transcript back to main.py where it is added to the user_prompt. Refactoring needed. Also since audio cuts off maybe better to send back the audio instead of the interruption idk.