Tetrak¶
Local-first OCR for archival material: postcards, posters, playbills, programmes and trade papers. Point it at a folder and it runs several engines per file, scores what each one produced, keeps the best, and quarantines anything it could not read well enough to trust. Everything runs on your machine unless you explicitly ask for the Claude backend.
This site is the user guide and reference for the tetrak package: how to
install it, how to drive it from the command line and from Python, and what
every command, flag and function does. It is built from the released code at
each release, so what it describes is the version on PyPI.
The research behind the tool lives on tetrak.dev: the evaluation corpus, what each engine was measured to do and where it fails, the evidence behind the routing, and the articles.
In two minutes¶
pip install 'tetrak[qa]' # the package, plus quality scoring
brew install tesseract poppler # macOS; see Installation for Linux
tetrak-ocr backends # what this machine can run
tetrak-ocr ocr scan.jpg # one file, to stdout
mkdir -p workspace/scans
cp ~/archive/*.jpg workspace/scans/
tetrak-ocr batch --backend auto-local # a folder, best engine per file
Transcripts land in workspace/processed/ beside the originals; anything too
poor to trust goes to workspace/triage/ with a note saying what was tried.
The quick start walks through that output.
Getting started
Reference