Implement audio recording for narration in record.py

[abrichr]

We would like to let the user explain what they are doing as they do it when creating a recording.

This is a replacement for https://github.com/MLDSAI/OpenAdapt/issues/57 , which is proving to be more difficult than we would like.

[FFFiend]

Do we want the recording to be captured and converted to be then mixed into the inputs for LLM completions somehow? (Whisper)

[abrichr]

@FFFiend exactly right!

[FFFiend]

Awesome, I'm just confused about one thing: once voice is captured, how exactly do we want to transform the data ? Perhaps a 1-1 mapping between Events and the actual content of the user's speech, because I couldn't see the reasoning behind generating completions for Audio when thats double the work that we're already doing. So is this more to increase the accuracy of completions (which is where the mapping comes in)? 😮

[abrichr]

@FFFiend good question. The idea here is to include the transcription in the model prompt to provide context to the model for what is important.

Related: https://github.com/OpenAdaptAI/OpenAdapt/issues/564