The website of OpenRAC, an unofficial hub for community decompilation projects of the Ratchet & Clank series. It shows, for every title, how far its decompilation has come.
Next.js (App Router) · TypeScript · Tailwind CSS 4 · everything is rendered on the server.
The page is a server component. On the server it reads the state of each project from GitHub, renders plain HTML and caches it; the browser gets finished HTML plus a few small scripts for the menu and animations. No visitor ever calls GitHub. The page is rebuilt in the background at most every 5 minutes, the numbers are cached for 10 minutes.
Every project is shown with one and the same measure: code matched, the bytes of code verified identical to
the retail build out of all code bytes the project counts. The percentage on each card and the line under its bar
(838 628 of 3 712 808 code bytes matched) are both this number; no card shows a project-specific extra. It is
weighted by size because small functions are matched first, and it is the figure decomp.dev shows.
Sources (see src/lib/progress.ts):
- Ratchet & Clank (PAL):
progress/report.json(objdiff report) ofOpenRAC/rac1-decomp. - Going Commando:
progress/report.jsonandconfig/progress-scope.jsonofOpenRAC/rac2-gc-decomp, branchRAC2. The percentage uses integrated C bytes over all executable bytes in the boot and 27 level overlays. Placements are occurrences across programs, not unique algorithms. The scope lists ELF sections, so it cannot supply a total function count; that denominator stays unknown. Exported objdiff unit counts remain report units, not source-file or function totals. - Up Your Arsenal: the objdiff report of
OpenRAC/rac3-uya-decomp, read from decomp.dev (/OpenRAC/rac3-uya-decomp.json). Its CI sends the report there on every push and does not commit it; theprogress_report.jsonin the repository is an old copy. - Ratchet: Deadlocked (NTSC-U):
progress/report.json(objdiff report) ofOpenRAC/rac-deadlocked-decomp. The project is in its research phase, which its card says; it is measured like the others. - Going Mobile: the verified method count in the
README.mdofClank700/going-mobile-decomp, applied to the size of the reconstructed source.
If GitHub cannot be reached, the last known snapshot (in progress.ts) is shown and labelled as such.
We use AI ourselves, so we do not block crawlers; we make the site easy to read instead. All of it is built from
the same data as the page (src/lib/machine.ts), so it never says anything the page does not:
| Address | What |
|---|---|
/progress.json |
every project and its numbers as JSON, open to any origin (CORS) |
/llms.txt |
a short Markdown map of the site for AI tools (llmstxt.org) |
/llms-full.txt |
the same, plus the FAQ and every blog post in full |
/robots.txt |
allows everything; Content-Signal: search=yes, ai-input=yes, ai-train=yes says so explicitly |
The home page and every blog post also carry schema.org JSON-LD (the projects as SoftwareSourceCode, posts as
BlogPosting), and the page head links the JSON, RSS and llms.txt versions.
The blog lives at /blog. Posts are Markdown files in content/blog/, added through pull requests:
---
title: A short title
date: 2026-09-30 # YYYY-MM-DD
summary: One or two sentences for the list page and link previews.
tags: [news] # optional
author: Name # optional
draft: true # optional, hides the post from the published site
---The file name is the address (my-post.md becomes /blog/my-post). A bad header fails the build and names the file.
Raw HTML in posts is ignored and images must be files in public/blog/ (remote images are dropped), so a post
cannot load anything from another server. There is an RSS feed at /blog/rss.xml, plus sitemap.xml and robots.txt.
src/lib/credits.ts lists the people credited on the front page and their YouTube videos (links or video ids).
The section stays hidden until a video is listed. Titles and thumbnails are fetched by the server, and a video
loads from YouTube (privacy-enhanced youtube-nocookie.com) only when a visitor presses play, so nothing
reaches Google before that.
npm install
npm run dev # http://localhost:3000
npm test # unit tests of the progress parser
npm run lint
npm run buildCopy .env.example to .env.local to change settings. GITHUB_TOKEN is optional and only raises the GitHub API
rate limit for the activity feed; it stays on the server.
The title cards can show a backdrop image named rac1-bg.webp, gc-bg.webp, uya-bg.webp, deadlocked-bg.webp and going-mobile-bg.webp (WebP, about
1600 px wide). Those would be screenshots from the games, which are copyrighted, so they are not part of this
repository. Without them the cards use a colour gradient.
The images are read from a folder, set by IMAGE_DIR (default public/img, which is git-ignored), on every
request. Drop a file in and it is used at once, with no rebuild or restart. Only those five names are ever
served. Resized copies are cached for a day, so a replaced image can take up to a day to show (or restart
the service after clearing the cache volume).
Add an entry to PROJECTS in src/lib/projects.ts. If the project publishes progress, add a loader to
src/lib/progress.ts; a title without one is shown as "Not started". maintainers (GitHub user names) puts
"maintained by" with their pictures on the card. Another project for a game that is already listed can be named
in the FAQ without a card of its own: add it to ALSO_WORKED_ON.
phase: "research" marks a project that is still mapping its game (shown on its card and in the FAQ; its
progress is tracked as usual). contributing and guide link the project's own contributing guide and best
introduction for newcomers; they feed the "Contribute" buttons and the "Project guides" under Resources.
# 1. put this repository on the server and build the image
podman build -t localhost/openrac:latest .
# 2. install the unit (rootless shown; rootful: /etc/containers/systemd/)
mkdir -p ~/.config/containers/systemd
cp deploy/openrac.container ~/.config/containers/systemd/
# 3. start it
systemctl --user daemon-reload
systemctl --user start openrac.service
loginctl enable-linger "$USER" # once, so it starts at boot without a login
# 4. card backdrops: copy your images into the folder the unit mounts (no restart needed)
cp rac1-bg.webp gc-bg.webp uya-bg.webp deadlocked-bg.webp going-mobile-bg.webp ~/openrac-img/The container listens on 127.0.0.1:3000; put your reverse proxy in front of it. It runs as a non-root user with
a read-only filesystem and all capabilities dropped; only the page cache (/app/.next/cache) is a writable
volume. The backdrop folder ~/openrac-img (created by the unit) is mounted read-only at /data/img; change the
host path in the Volume= line if you keep the images elsewhere. A health check polls /healthz. Edit SITE_URL in the unit to match your domain.
To update: git pull && podman build -t localhost/openrac:latest . && systemctl --user restart openrac.service.
OpenRAC is an independent fan project, not affiliated with or endorsed by Sony Interactive Entertainment or Insomniac Games. "Ratchet & Clank" and related names are trademarks of their owners and are used only to identify the games the linked projects study. This repository contains no game assets, code or binaries.