jolly sky
Hey everyone,
Lamine here, I'm a dev and I built a small tool called DocLime, essentially ChatGPT for PDFs.
As simple to use as chatgpt.
New conversation -> Drop your PDF or choose from library -> Start asking questions
I made a free version no credit card required. (5 questions)
If you guys want to give it a try! I'd love to hear your feedback,
Cheers,
Lamine
kind oracle
Will give it a try!
prime tartan
How well does it handle scanned in PDFs? Are you using pdfplumber or another method?
Ill assume that you are embedding the text from the PDF and then using cosine to match similarity? Or are you doing sentiment analysis with turbo?
junior burrow
100000% will try. Great feature man
signal gazelle
Would like to try it out
restive granite
wow!! it’s so useful thank you for this!
jolly sky
Will give it a try!
Thank you
jolly sky
How well does it handle scanned in PDFs? Are you using pdfplumber or another me...
It doesn't handle scanned PDFs at the moment. That'll come soon.
jolly sky
100000% will try. Great feature man
Thank you!
jolly sky
Would like to try it out
Thank you! appreciate it
jolly sky
wow!! it’s so useful thank you for this!
Looking forward to get you in the app, cheers
restive granite
Looking forward to get you in the app, cheers
sounds good!! thank you so much for this!!
prime tartan
Scanned PDFs would be incredible, but my understanding is that would need ocr, get to it. Lol.
grave marlin
@jolly sky How can I test your app?
jolly sky
sure you can try it : https://doclime.com @grave marlin
grave marlin
sure you can try it : https://doclime.com <@243056506419806208>
thanks bro
frozen thunder
nice work! @jolly sky do you have any guides on how you created this?
junior burrow
Definitely will give a try. I've built a bunch these myself in different languages, using different methods, the works and I get close but then you get to that one easy question and it just says "Elon Musk" or something. @jolly sky the biggest hurdle, I think for everyone who has tried to develop a tool like this (and you mention it on your website) is figures. For myself, I can never get it to figure out how to handle financial statement figures (what I mostly use as test subjects) where it doesn't eventually mess up in some way. You say it's in the works, but how close are you to achieving this as I know it would make a very large community happy. Love the architecture of the project though. Are there any tools that exist that has been able to tackle this yet? It's talked about so much but none of us can solve it. Don't worry, if you do know of a tool that can handle financial data, im still signing up for a paid version to try of your platform 🧙♂️
spiral nexus
I was asked to upgrade to Starter. No scanned PDF reading ability as yet? Is this running on 4 or 3.5?