Project

DocScan: a self-hosted document scanner

Any phone, any browser, no app install: photograph a page and get back a clean, deskewed PDF, without a third-party scanning app ever seeing the document.

The problem

Paper still piles up: receipts, mail, forms. The quickest fix is one of the free phone scanning apps, but those are built to see (and often keep) whatever you photograph. I wanted the same one-tap convenience without handing every document I own to someone else's cloud.

What I built

DocScan: a web app I run myself at scan.sh-nxn.dev. Open it on any phone, capture a photo, and a computer-vision microservice detects the page's corners automatically (with a manual nudge if it guesses wrong). The result gets cropped, deskewed and turned into a proper PDF, then pushed straight into my document archive, no manual filing step.

It runs as two small services behind one login: the scanning app itself, and a separate image-processing service it talks to internally. Neither one is reachable from outside except through the scanning app's own login.

Tech

Node.js / Express
FastAPI
OpenCV
SQLite
Paperless-ngx

DocScan's sign-in screen on a dark background, with the app name and a 'sign in to start scanning' prompt — What you see first: no account self-signup, sign-in only, same as every other service in my homelab.

"Most of the real work wasn't the scanning, it was deciding what the app should never do on its own: it never guesses a document is 'done', it never deletes the original photo until I confirm the PDF looks right."