paperextract

Turn a folder of scientific PDFs into a searchable, citable, machine-readable library, locally, without losing the equations, tables or figures.

paperextract runs document models on your own machine or cluster, keeps every result traceable to the page it came from, and reports what it could not verify instead of guessing.

uv run paperextract paper.pdf --library ~/papers
uv run paperextract search "soliton self-compression" --library ~/papers

What you get

• Equations, tables and figures as data: LaTeX math, exact table cells as HTML, JSON and CSV, figures with panels and captions, supplements linked to the main text.

• Nothing invented: original files kept with hashes, values never repaired, every doubt recorded as a finding.

• Validated identity: DOIs, titles and authors checked against Crossref, DataCite and arXiv, with BibTeX and document versions.

• Duplicates caught first: identical files, same-text copies and shared DOIs found before any model runs.

• A searchable, portable library: full-text search, lookup, library.bib and an agent skill for AI assistants.

• Figure descriptions: optional, labelled machine descriptions.

• Laptop to cluster: MinerU, Docling or Marker on a Mac or Linux, and GPU batches on Slurm.

Where next

• Getting started: install, extract a first paper, see what it produces.

• Worked example: one real paper, input against output.

• Command line: every command and setting.

• Running on a Slurm cluster: GPU batches.

• Known limitations: read before relying on a result.