antifile

Extract BibTeX metadata from PDFs, EPUBs, URLs, and identifiers (DOI, arXiv, ISBN) and append it to a .bib file.

Install

pip install antifile
# or
uv tool install antifile

Usage

antifile INPUT -o refs.bib

INPUT can be:

• a PDF or EPUB file — antifile paper.pdf -o refs.bib
• a folder of PDFs/EPUBs — antifile ~/Downloads/papers -o refs.bib (add --recursive to descend into subfolders)
• a URL — antifile https://example.com/article -o refs.bib
• a DOI — antifile 10.1145/3292500 -o refs.bib
• an arXiv ID — antifile arXiv:1706.03762 -o refs.bib
• an ISBN — antifile 9780262033848 -o refs.bib

Entries are appended with de-duplication: a new entry matching an existing one (by DOI, arXiv ID, ISBN, or normalized title+author) fills in any missing fields rather than creating a duplicate. Pass --no-merge to skip on duplicate, or --force to append anyway with an auto-suffixed key.

Options

LLM-assisted extraction

When a PDF has no resolvable identifier, antifile falls back to an LLM to read the first page. Set whichever API key you have — it's picked up from the environment:

export ANTHROPIC_API_KEY=...   # or OPENAI_API_KEY, or GEMINI_API_KEY / GOOGLE_API_KEY

With no API key set, antifile falls back to a CLI agent if one is installed — first the claude CLI, then the codex CLI. Each tier is skipped if its binary isn't on your PATH. Full chain:

DOI → arXiv → ISBN → CrossRef title → LLM API → claude CLI → codex CLI → none

Related

• antilibrary — manage BibTeX libraries from the terminal (can call antifile via --add-from-files).
• antifind — online metadata search → BibTeX.

License

MIT