Add pdfmux to PDF processing tools

[NameetP]

Adding pdfmux under PDF processing tools.

What it is

A Python orchestrator that classifies each PDF page (digital text, scanned, tables, mixed) and routes it to the optimal backend — PyMuPDF for digital text, Docling for tables, RapidOCR for scans, Gemini Flash for hard pages. Emits Markdown plus a per-page confidence score so document understanding pipelines can quarantine low-trust pages instead of feeding noise to KIE / LIR / classification models downstream.

Why it fits this list

• Covers the messy ingestion step that precedes the KIE/LIR tasks this list focuses on. pdfmux doesn't do KIE itself; it produces clean Markdown + confidence signals that feed into the information extraction work downstream.
• Confidence scoring is the differentiator. Rather than a binary "did it extract or not?" the score lets IE pipelines route low-confidence pages to human review or re-extraction.
• Sits next to peers already in this section: borb, pdfplumber, pdfminer.six, Layout Parser, deepdoctection.
• Open source (MIT), on PyPI: pip install pdfmux. Active maintainership — v1.5.0 released April 2026.
• Benchmarked on 1,422 pages across 11 real-world business documents (10-Ks, S-1s, academic papers, legal opinions, FDA reports) with 100% confidence retained.

Links

• Repo: https://github.com/NameetP/pdfmux
• Docs: https://pdfmux.com
• PyPI: https://pypi.org/project/pdfmux/
• Benchmark methodology + per-doc results: https://pdfmux.com/blog/real-world-test/