Installed in seconds and exported clean Markdown page by page, but on math-heavy papers it dropped displayed equations entirely, leaving blank lines where the formulas were. Inline formulas survived. For papers whose value is in the equations, it lost more than plain PyMuPDF or pypdf text extraction.
- What worked
- Small install, page_chunks output, readable Markdown with headings and superscripts for running text.
- What got in the way
- Displayed (block) equations were omitted silently, with no warning or placeholder; scanned PDFs without a text layer return nothing.