Inspiration
If you're part of an extremely small agency like we are, you'd know that scanned documents are a fact of life wherever physical documents are concerned. Whether they be contracts, receipts or proofs, these are usually scanned before the signing process starts. Ironically this scanning process is the hardest to get right, with mobile phones being used by local vendors (since it's quick and convenient)
What it does
A PDF reader that takes a raw scanned PDF and converts it into a ready to sign document, then uses the Dropbox Sign API to send it out for signing. It does all this while integrating to the GPT api as well as implementing a custom model that segments and aligns documents, in order to make it easy to construct an email on the fly for signing. Very simple, very secure, and very easy
How we built it
An Electron app using onnx.js (with a custom Pytorch model) for scanning the documents and tessaract.js for OCR. Uses ChatGPT for summarizing documents and writing email messages and Dropbox Sign API to request for a signature.
Challenges we ran into
Using pytorch models in javascript was hard, as was calling ChatGPT from the client side, but overally the Dropbox Sign API seemed decent enough to use, and was pretty easy to integrate
Accomplishments that we're proud of
Working hard on an electron app that could be integrated into our own daily workflow was a highlight of the hackathon for sure
What we learned
Definitely how to use the Dropbox Sign API
What's next for Monsigneur
Develop it further to add more features
Built With
- chatgpt
- onnx
- pytorch
Log in or sign up for Devpost to join the conversation.