tetrak_ocr¶

Batch OCR pipeline and backend comparison harness for archival images.

Every backend exposes ocr_image(path) -> str. Resolve one by name through tetrak_ocr.registry rather than importing it directly, so that a missing optional dependency reports which extra to install.

>>> from tetrak_ocr.registry import get_backend
>>> ocr = get_backend("tesseract")
>>> text = ocr(Path("scan.jpg"))

Modules

`accuracy`	Text similarity metrics for OCR evaluation.
`auto_local`	auto-local OCR backend: fan out over the local engines and keep the best.
`backends`	OCR backends.
`batch`	Batch OCR processor.
`cli`	Command line entry point.
`errors`	Exceptions raised by the package.
`imaging`	Frame handling for multi-page raster images.
`layout`	Positioned text: the vocabulary shared by the backends and the PDF writer.
`outputs`	Output formats for a finished transcript.
`pdf_output`	Write a searchable PDF: the scan, with the transcript invisible over it.
`qa_score`	Reference-free OCR quality metrics.
`registry`	Resolve a backend by name.
`telemetry`	Append-only run log: what was processed, by which backend, and how long it took.