document2md¶
Version 0.4.0. document2md converts a PDF, or an
ordered set of scanned page images, into Markdown, via one of two backends
— mineru (OCR and layout
analysis) or pymupdf4llm (reading a
born-digital PDF’s own embedded text layer, no OCR needed) — optionally
cropped down to a single section. This package was extracted from the
LegalIA monorepo at commit
e1f258c; how it fits with the packages it used to share a repository
with is covered on the user site.
Note
This is the developer documentation: architecture, backends, and the full API reference, for extending the package itself. If you only want to use document2md — install it, run it from the command line or from Python, pick a backend — see the user site linked above.
How this documentation is organised¶
- Architecture
document2md.cli— command-line entry pointdocument2md.batch— Python entry pointdocument2md.mineru_server— keeping mineru-api warmdocument2md.converter— running minerudocument2md.pymupdf_backend— reading a PDF’s own text layerdocument2md.tables— HTML tables to Markdown tablesdocument2md.cutter— cropping to a single section
- Backends
- Development
- API reference
Architecture walks through the conversion pipeline end to end, module by module, in the order a document actually flows through it. Backends explains how backend resolution and dispatch work, and how to add a new backend. Development covers the repository’s own conventions — tests, docs and website builds, versioning, publishing. API reference is the full API reference, one page per module, including private/internal helpers — useful when extending or debugging the package, though they are not part of its public API and can change without notice.