Waraqi 0.2.0-beta.2

Waraqi 0.2.0-beta.2

Second release. PDFs, a settings screen, a PIN lock, editable text, and an
optional AI assistant that asks before anything leaves your machine.

New

PDF import. Drop in a PDF and every page is read. Pages are checked one at a
time rather than the file as a whole, so a page that already has a real text
layer is read directly and skips OCR entirely. Only scanned pages go through
recognition, which makes text based PDFs import almost instantly. Up to 30 pages
are recognised per file, and the viewer renders up to 20.

Editable extracted text. OCR gets things wrong. Fix it in the viewer and the
search index updates immediately.

Tag sidebar. Every tag with a count, clickable to filter. Tags are matched
through the same Arabic normalization as search.

File format filter. A dropdown next to search filters the library by format,
listing only the formats you actually have with a count for each.

Settings screen. Theme, interface language, PIN, and the AI configuration in
one place.

PIN lock. A four digit PIN asked for when the app opens. Read the limitation
below before you rely on it.

Optional AI assistance. Off by default, runs on your own Gemini key. Two
actions: clean up the extracted text, or re-read the image. Both show a
confirmation naming the third party, and for images the size of the upload, with
a "do not ask me again" option that can be reversed in Settings.

Rename documents, and open the library folder from the app.

Installers for Windows, (macOS and Linux installers will be released later).

Known limitations

• The PIN locks the window. It does not encrypt your documents, which stay
  readable in the library folder to anyone with access to the machine.
  Encryption at rest is the next piece of work.
• Accuracy drops on low resolution photos and on documents printed over security
  patterns. This is why the extracted text is editable.
• Automatic typing only knows eight document types. Anything else imports
  untagged and you can tag it by hand.

Upgrading from 0.1.0-beta.1

The database schema changed. If the app reports that your database is from an
older version, it will name the exact file to delete. Your documents in the
library folder are not touched, but you will need to import them again.

Install

Windows: Waraqi_0.2.0_x64_en-US.msi, or Waraqi_0.2.0_x64-setup.exe.

The build is unsigned, so Windows SmartScreen will warn you. Choose More info,
then Run anyway.

Waraqi 0.1.0-beta.1

First release. An offline document archive that reads Arabic and English.

Photograph your documents, Waraqi extracts the text, and you can search it later.
Everything runs on your machine. There is no account, no server, and the app makes
no network requests at any point, including the first run.

Install

Windows 10 or 11

• Waraqi_0.1.0_x64.msi
• Waraqi_0.1.0_x64-setup.exe

Not code signed, so SmartScreen will warn you. Click "Run anyway".

macOS

• Waraqi_0.1.0_aarch64.dmg (M1 and later)

Not signed or notarized, so right-click the app and choose Open the first time.

Linux

• waraqi_0.1.0_amd64.deb (Debian and Ubuntu)

Windows is the build we tested. The macOS and Linux builds are produced by CI on
their own platforms and pass the test suite, but we have not been able to run them.

You can build any of these yourself from source if you would rather not trust ours.

What is in it

• OCR in Arabic and English together on every page, including mixed documents
• Full text search in either language, with Arabic normalization so فاتوره finds فاتورة
• Tags, searchable, with counts in the sidebar, applied automatically from the page
• Rule based document type detection, no model involved
• Export a document as a searchable PDF, or the whole archive as files plus a CSV
• Arabic and English interface, right to left and left to right, dark and light

Where your documents live

Windows: %APPDATA%\ngo.josa.waraqi\
macOS: ~/Library/Application Support/ngo.josa.waraqi/
Linux: ~/.local/share/ngo.josa.waraqi/

That folder holds the imported images and a single SQLite database. Nothing leaves
it. Delete it and the app is empty again. The archive export exists so you can take
everything with you in a form that opens without Waraqi.

Known limits

• Images only, multi page PDF import is not supported yet
• OCR quality depends heavily on the photo. Fill the frame, avoid shadows across the text
• The invisible text layer in exported PDFs is searchable but not correctly shaped Arabic

Built by

Viscanturtle and odytr for Reclaim Hackathon by JOSA.

MIT licensed. Third party components are listed in THIRD_PARTY.md.