.gitignore

_site
.sass-cache
.jekyll-metadata
Gemfile.lock
__pycache__/
*.pyc

# Impact pipeline caches. The bulky, disposable layers -- fetched HTTP bodies,
# GitHub payloads, scholarly API responses -- are re-fetchable and restored from
# the GitHub Actions cache instead. The compact layers below them are committed
# because they are the collector's reproducible state: what has been scanned,
# what was deterministically extracted, and every classification answer.
.cache/impact/http/
.cache/impact/github/
.cache/impact/papers/
.cache/impact/artifacts/
.cache/impact/last-run.json
.cache/impact/pr-body.md
.jekyll-cache

# Hand-supplied paper PDFs. These stay on the machine that downloaded them:
# they are publisher copies, often licensed to one reader, and the pipeline
# only ever needs them locally. Everything in the folder is ignored except the
# two files that explain it, so a PDF cannot be committed by arriving with an
# unexpected name -- an uppercase .PDF, which the loader also reads, or a
# subfolder of a bulk download.
.cache/impact/papers-pdf/*
!.cache/impact/papers-pdf/README.md
!.cache/impact/papers-pdf/WANTED.md

# Extracted paper text, cached locally. The dossiers in _data/papers keep the
# mentions and their context; the whole text of a publisher's PDF does not
# belong in the repository.
.cache/impact/papers-text/