Release 0.0.22 has been published

[JorjMcKie]

Apart from fixing some issues, the major improvement is the new class TocHeaders.

It can be used as an alternative to the standard header recognition class IdentifyHeaders, which relies on a full document text scan to find popular font sizes and uses them for identifying headers.

For documents with a well-formed, detailed TOC the new class is much better:

• More precise (as precise as the TOC itself):
  • Bookmark levels directly translate to header levels
  • No false positives
• Faster because no preliminary text extractions are performed

Typical usage code snippet:

import pymupdf, pymupdf4llm, pathlib

doc = pymupdf.open("input.pdf")
toc_headers = pymupdf4llm.TocHeaders(doc)

md = pymupdf4llm.to_markdown(doc, hdr_info=toc_headers)

pathlib.Path(doc.name).with_suffix(".md").write_text(md)