FMJL

FMJL, short for Format, Markdown, JSON Lines, stores a document as one JSON object per line: one element (heading, paragraph, table, formula, image, caption...) with a stable id, a content hash, its page and position, and who may read it. People write and edit the same document as normal Markdown; the tool converts both ways without renumbering anything.

pip install fmjl            # converter, checker, Word importer, chunks
pip install fmjl[pdf]       # adds the PDF importer (PyMuPDF)

fmjl notes.md               # Markdown  -> notes.fmjl, then checks it
fmjl notes.fmjl             # .fmjl     -> notes.md, ids kept
fmjl report.pdf             # PDF       -> report.fmjl and images/ (--md adds report.md)
fmjl report.docx            # Word      -> the same
fmjl chunks report.fmjl     # retriever-ready chunks as JSON Lines
fmjl export report.fmjl --to docx   # also md, html, pdf, odt, epub

From Python: import fmjl; fmjl.load, fmjl.chunks, fmjl.changed_chunks, fmjl.check.

Rulebook, examples, benchmark and the VS Code extension: https://github.com/rupakkumar-76319/fmjl