Reference-list extraction from PDF and DOCX

[chrisyangsong]

Today citegate only reads BibTeX. Reviewers and editors usually receive a PDF or DOCX. Add an extraction front-end that pulls the reference list out of a PDF/DOCX (e.g. GROBID or a lightweight regex/anystyle approach) and feeds parsed entries into the same verification pipeline.

Design question to settle first: bundle extraction as an optional extra (pip install citegate[pdf]) so the core stays dependency-light.

[fasuizu-br]

Affiliation: I am writing on behalf of BRAINIALL, which operates a metered hosted text-layer PDF-to-Markdown route and does not provide OCR or bibliography parsing. I will not open a PR, request credentials, or upload private or non-rights-cleared documents. Since #3 leaves the PDF extraction front-end and optional dependency boundary open, would one maintainer-selected synthetic two-page text-layer PDF containing a short rights-cleared reference list plus one image-only control be useful as a bounded external observation, with fixture hashes, page-range request, returned Markdown or explicit empty behavior, response MIME, elapsed time, and observed charge posted here? This would not claim DOCX, GROBID, regex or anystyle parsing, citation verification, semantic fidelity, layout, OCR, dependency packaging, or citegate integration.