docext

Why we included this project

docext is a sensible pick for teams that need structured fields out of invoices, passports, receipts, and long scanned PDFs without standing up a heavy OCR stack. It is an on-premises toolkit that uses vision-language models to convert documents and images to markdown, picking out tables, LaTeX equations, signatures, watermarks, and form checkboxes along the way. The same extraction pipeline returns fields with confidence scores, so a database or retrieval system gets clean, queryable output instead of raw OCR text. A bundled benchmarking harness also lets you compare how different models handle key information extraction, table parsing, and document classification on your own data before you commit to one.