todo.txt

# TODO:
- Wake word voice activation
- Drop down for TTS voice choices
- Clean UI (Text at bottom, better icons and stylizing)
- Hide API key✔️:check / Allow user to input their API key
- Seperate functions for Image Chat and just text calls
- chunk audio over 30 seconds for whisper
- Whole chat context for conversation 
- UI similar to my vercel app / chatgpt
- a settings widget in bottom corner to change voice / and display settings
- Error Feedback to Users:
- input validation
- preformance optimization
- payments
- authentication 
- Streaming TTS and Streaming Text
- Window Selection for live and ability to exit the screen share.
- Audio source detection / choose source

- Rename variables / button and delete unused code. 

Ignore this Image

Main task for 11/14:
    Get message history working so old messages from api are included in context✔️
    Option to clear the message history to start fresh chat✔️
    Streaming text completion? ✔️

Main tasks for 11/15:
    Easier:
        Get the response UI formatted properly, input at bottom like chat apps ✔️X sorted
    Challenge:
        Get the streaming TTS api working, this will be quite involved as it will require passing chunks✔️
        of the audio to the app similar to the text ✔️ still funky
    bonus:
        get a window detection drop down working
        user selects window to stream 
        ability to stop stream

Main tasks for 11/16:
    Get TTS Pause, Skip, Start Stop working✔️ 
    Also prepare self for big react refactor with vercel chat-ai template as inspiration  

Main Tasks for 11/17:
    Lets get window streaming working✔️ 
    Select window to stream when live:✔️ 
    send window along with prompt to image api✔️ 
    ability to stop the stream and go back to normal prompts✔️ 
    bonus:
        settings functions for api key, window settings, and voice actor settings

We have to fix API call logic if no image is included we can't use the vision api
i think its more expensive and their is a small daily limit

Idea:
Set a live window > calls be seconds > and just return an image call on the interval:
basically lets you 'listen' to your frame/window instead of watching it


BUGS: Seems like the Live function fights with the snip tool, might have to handle if snip ignore live or something
TTS only seems to work when I use the record feature look into this
Capture is still working after toggled off
Snip tool breaks the capture function it works in isolation but breaks snip