- Look into why it takes so long at times
- Look into weird \ tokens
PAUSED UNTIL FAST JSON MODE IS OUT.
- Newlines for summaries
- Reject PDFs that are too large or over X pages (maybe 1k?)
- Try out finetuned models (likely won't go with them)
- Don't have a PDF? Try out this one w/ synthetic loading
- Add OG image & metadata + analytics
- Add new S3 bucket for Smart PDFs
- Maybe add more parsing to make sure its JSON
- Update links throughout app to be dub links
- If there are many sections, combine them into one
- Add a thumbs up and thumbs down feature
- Try out additional revision steps to make the summary even better
- Use OCR to parse out any images in the PDF – test this out: https://menlovc.com/2024-the-state-of-generative-ai-in-the-enterprise/