jcrt-files // files.jcrt.org #

Asset delivery repository for JCRT.

TODO (Adam) #

• Grant the CLOUDFLARE_API_TOKEN Workers deploy permission. Resolved 2026-08-30: after the re-paste fixed the formatting and the permission grant added Workers Scripts: Edit, run 33297892081 logged Deployed with TOKEN_PRIMARY. The deploy-is-the-test token fallback (PRIMARY → V2 → BACKUP) stays in place as insurance.

Purpose #

This repo receives synced assets from jcrt-v2 and publishes them to Cloudflare R2, served by a Worker on:

• https://files.jcrt.org/images/...
• https://files.jcrt.org/archives/...
• https://files.jcrt.org/citations/...
• https://files.jcrt.org/docs/...

Required secrets (GitHub repo: jcrt-files) #

• CLOUDFLARE_API_TOKEN
• CLOUDFLARE_ACCOUNT_ID

Required Cloudflare setup #

1. Create R2 bucket: jcrt-files.
2. Deploy worker from this repo (wrangler deploy).
3. Add Worker custom domain route for files.jcrt.org/*.
4. Ensure DNS for files.jcrt.org is in Cloudflare-managed jcrt.org zone.

How sync works #

• jcrt-v2 workflow pushes mirrored content into this repo.
• Every push runs an rclone --checksum comparison of the whole git-tracked tree against the bucket and uploads whatever differs (content-addressed — no diff windows, so a file committed during a cancelled run is picked up by the next one). The git tree is the source of truth.
• Deletion is opt-in: dispatch the workflow with prune_deletes=true to remove bucket objects no longer tracked in git.
• The Worker deploys only when src/ or wrangler.toml changed (or deploy_worker=true).