Add OpenAI ChatGPT compatible integration

[stevenlafl]

[abb128]

FYI the flatpak app doesn't have the network permission so this wouldn't work without adding the network permission now, and it would kinda violate the offline and privacy-first aspects of the app. I will likely not be merging this but you're welcome to maintain your own fork with these added online features

[stevenlafl]

It might be a good start for a local model - though I am not familiar enough with CPU-only ones. I'll think about how that might work.

My use-case is to transcribe meeting notes and then be able to ask questions about it, or generate summaries. A Krisp-like setup is the desired end-state.

[abb128]

Do you think it might be better to improve copy-pasting transcripts instead, to better facilitate pasting it into ChatGPT or the user's assistant of choice?

[stevenlafl]

Short answer is yes, talking about the aims and scope you mention associated with this project.

Ideally I'd need it to:

1. Start/stop "sessions" on demand and (optionally) automatically based on a silence-for-time period threshold. I have a lot of back-to-back meetings, and I may simply forget to restart the app.
2. Copy transcripts out easily, optionally with timestamps (perhaps a checkbox?). A "copy latest session transcript to clipboard" button would work well.

The reason I did it this way is that sometimes I get distracted and need some parts summarized, to get context quickly and easily while a conversation is actively happening. However, it is certainly still doable with a copy button.

Optionally, not sure how possible:

1. Speaker diarization, but this is probably more for the april-asr project and may be far, far out of scope. A simple '>' as indicated on actual live TV transcripts denotes when a new speaker starts talking. The "voice print" just needs to change enough to trigger that, not actively track how many speakers.