[Paururo]

The documentation was fourteen MkDocs pages and 23,000 words. This replaces the site with two pages meant to get someone running in a couple of minutes, and keeps the long versions as markdown on GitHub.

The website (docs/index.html and docs/reference.html: plain HTML, CSS and JavaScript, no build step, nothing loaded from another host)

• The landing page opens on an animated matrix that shows what Recore does: the empty cells are searched, eight genes are put back, three are kept out with the reason and twelve stay absent. Then:
  1. Install: minimap2 (conda, Homebrew or apt) and cargo install --git https://github.com/PathoGenOmics-Lab/recore --locked.
  2. Your command: pick Panaroo, Roary, PPanGGOLiN, PanTA or PIRATE and tick what you have (GFFs, a tree, a reference genome, the HTML report). The command, a line-by-line explanation and the folder layout it expects update as you go. The file names per tool are the ones docs/getting-started/inputs.md checked on real output.
  3. Try it first: three commands on the tutorial dataset and the counts they give (Rescued 11, Matrix-corrected 8).
  4. What you get: the four files that answer most questions.
  5. The example report, embedded. It only takes the mouse wheel after a click, so the page keeps scrolling; phones get a full-screen link instead of the frame.
  6. The one setting that matters: the three matrix policies.
  7. Seven first questions.
• The reference page has every option (grouped, with defaults, and a filter), every output file, the thirteen cell states in the report's colours, the rescue types, and links to the long pages.
• Light and dark themes, visible keyboard focus, reduced motion respected, no horizontal scroll at 375 px.

Kept honest in CI (tools/check_site.py, run by the Documentation / Build job; the names are unchanged for the Dependabot gate)

• every internal link and anchor resolves, and nothing loads from another host;
• every option the pages or the command builder show exists in recore --help, the reference lists every option, and its defaults are the binary's;
• no em-dashes;
• the "Try it" commands run as written, and summary.txt must give the counts the page promises.

Broken on purpose six ways (an unknown option in the builder, a wrong default, a dead anchor, an option missing from the reference, a remote stylesheet, a wrong promised count), it failed each time.

Removed and moved

• mkdocs.yml, requirements-docs.txt, the MkDocs home, installation, quick start and changelog pages, and the pip entry in dependabot.yml. The workflow no longer deploys: GitHub Pages can serve docs/ as it is (Settings, Pages, main and /docs) once the repository is public, and docs/.nojekyll keeps Jekyll out of it.
• The long pages lose their MkDocs-only syntax (admonitions, definition lists, ++key++ shortcuts) so they read properly on GitHub, and their links to removed pages point at the README.
• The tutorial no longer says the report has to be served over HTTP: opened from disk it starts in Chrome and in WebKit (checked headless, and with a WKWebView probe).

Checked by running it

• cargo install --git https://github.com/PathoGenOmics-Lab/recore --locked into a clean directory, then the Try-it commands with the dataset script on its own: Rescued 11, Matrix-corrected 8.
• Both pages looked at in light, dark and at phone width in a browser.
• fmt, clippy, all tests with minimap2 required, the auto-merge gate check, and the site check with the tutorial run.

The old gh-pages branch with the last MkDocs build is still on the remote and can be deleted.