Skip to content

Overview

Docudis replaces personal details in a document with numbered labels such as [PERSON_1] or [IBAN_1], on your own device, so you can give the text to an AI service. When the AI replies, Docudis puts the original values back.

  1. Open a document. Paste text, open a PDF, Word or text file, or take a photo of a paper document.
  2. Docudis finds personal details. Rules, word lists and an on-device name-recognition model look for names, emails, phone numbers, ID numbers, IBANs, addresses and more.
  3. You review the result. Every detection is shown. You can add or remove items before anything leaves the app.
  4. Send the protected copy. Copy the text, share a protected file, or send it straight to an AI app.
  5. Restore the reply. Paste the AI’s answer and Docudis swaps the labels back for the real values. The label list never leaves your device.
  • It can miss things. No automatic detection finds every personal detail. Read the protected copy before you send it.
  • It hides identifiers, not meaning. A diagnosis, a job title or a story can still point to a person.
  • Dates and amounts stay visible by default. They are detected but only hidden if you choose to hide them.
  • Restoring needs the labels. It tolerates small formatting changes, but a label the AI rewrites beyond recognition is not restored.

Everything is open source.

Part What it does License
docudis-android The Android app AGPL-3.0
docudis-desktop The Windows and macOS app AGPL-3.0
docudis-core Rules, word lists, merging, replacement and restoration, with a C interface Apache-2.0
docudis-ner Running name-recognition models Apache-2.0

Dependencies point one way: docudis-ner depends on Core, and the apps depend on both.