gha-rxiv-feed-action

gha-rxiv-feed-action is a GitHub Action that logs a weekly CSV feed of new arXiv, bioRxiv, medRxiv, and PsyArXiv papers for selected categories, for downstream paper-evaluation pipelines.

What

• Fetches new papers from the selected SERVER (arXiv by default; bioRxiv, medRxiv or psyArXiv on request)
• Filters by arXiv TOPICS or bioRxiv/medRxiv/psyArXiv CATEGORIES
• Writes one CSV per ISO week to ./data/<server>/<year>/<isoweek>.csv, deduped by (ID|DOI, Version)
• Carries title, authors, and abstract on every row; optional Semantic Scholar citation counts for arXiv, bioRxiv and medRxiv
• Backfills any historic window via DATE_FROM/DATE_TO
• Opens a PR with the updated data and auto-merges it via squash
• Cron cadence is set by the calling workflow

How

Default: fetch the last week of arXiv submissions in CS/AI categories.

- uses: qte77/[email protected]
  with:
    TOKEN: ${{ secrets.GITHUB_TOKEN }}

See docs/usage.md for bioRxiv/medRxiv, multi-server matrices, backfill dispatches, the full inputs table, and versioning; docs/data.md for the API sources and CSV schemas.

Why

Existing arXiv-digest bots post keyword matches as issues for humans to read, and cover arXiv only. This action instead persists a normalized, deduped weekly CSV across arXiv, bioRxiv, and medRxiv, so a separate evaluation step can consume one schema. See docs/landscape.md and docs/userstory.md.

Refs

• docs/usage.md — examples, inputs, versioning
• docs/data.md — API sources and CSV schemas
• docs/architecture.md — data flow and invariants
• docs/categories.md — per-server taxonomy reference
• docs/landscape.md — competitive scan of similar feed bots
• docs/roadmap.md — narrative summary; issues are the source of truth
• docs/userstory.md — who this serves
• CHANGELOG.md — changes

License

Apache-2.0 — see LICENSE.