It's kind of like Rewind but it barely works and it's mine!
Yes I know using Rust was a terrible idea.
- Enumerate windows and screenshot them
- Skip Incognito windows
- Diff windows to detect changes
- Generate some multimodal embeddings
- Generate better multimodal embeddings
- Store the embeddings
- Make a UI accessible from the status bar
- Query for best match embeddings from a UI
- Allow pausing/resuming since Incognito detection is probably impossible
- Do something with Vision OCR results?