docs: publish redirects to commit-check.com ahead of archiving

.github/workflows/deploy.yml

@@ -1,4 +1,14 @@
-name: Deploy Documentation
+name: Deploy redirects
+
+# This site moved to https://commit-check.com. What this repository publishes
+# is a set of redirect stubs, built by scripts/build_redirects.py rather than
+# by mkdocs.
+#
+# The deploy matters more than usual: once this repository is archived, Actions
+# stop running and Pages keeps serving the last artifact indefinitely. The
+# redirect map is therefore checked on every run — a URL left out is a link
+# that breaks permanently.
+
 on:
   push:
     branches: [main]
@@ -19,10 +29,20 @@ jobs:
         with:
           python-version: '3.x'
 
-      - name: Install Dependencies
-        run: pipx run nox -s docs
+      # mkdocs is still needed: the check builds the old site and compares the
+      # URLs it produces against the redirect map.
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install pytest -r docs/requirements.txt
+
+      - name: Check every old URL still has a forwarding address
+        run: pytest tests/ -q
+
+      - name: Build redirects
+        run: python scripts/build_redirects.py
 
-      - name: Upload docs build as artifact
+      - name: Upload redirects as artifact
         uses: actions/upload-pages-artifact@v5
         with:
           name: ${{ github.event.repository.name }}_docs

README.md

@@ -1,5 +1,47 @@
-# commit-check.github.io
+# commit-check.github.io — moved to commit-check.com
 
-[![Website](https://img.shields.io/static/v1?label=Website&message=commit-check.github.io&color=2c9ccd&logo=git&logoColor=white)](https://commit-check.github.io)
+> [!IMPORTANT]
+> **This repository is no longer the source of the Commit Check website.**
+>
+> The documentation, the landing page and the blog now live at
+> **[commit-check.com](https://commit-check.com)**, built from
+> **[commit-check/commit-check.com](https://github.com/commit-check/commit-check.com)**.
+>
+> Open issues and pull requests about the website there, not here.
 
-This repository contains the source code for the [commit-check.github.io](https://commit-check.github.io) website.
+[![Website](https://img.shields.io/static/v1?label=Website&message=commit-check.com&color=2c9ccd&logo=git&logoColor=white)](https://commit-check.com)
+
+## What this repository is now
+
+It is kept as the historical source of the old site. The `docs/` directory is
+still here and still readable, but it is no longer published.
+
+What `commit-check.github.io` serves is a set of redirect stubs — one for every
+URL the old site had — pointing at the page that replaced it. The pages, the
+blog and its archive, author and category indexes all kept their paths, so the
+redirects are one-to-one; `/projects/` is the exception, having been folded into
+the Ecosystem section of the new landing page.
+
+| Old URL | Now |
+|---|---|
+| `commit-check.github.io/` | [commit-check.com/](https://commit-check.com/) |
+| `commit-check.github.io/getting-started/` | [commit-check.com/getting-started/](https://commit-check.com/getting-started/) |
+| `commit-check.github.io/blog/…` | [commit-check.com/blog/…](https://commit-check.com/blog/) — same paths |
+| `commit-check.github.io/projects/` | [commit-check.com/](https://commit-check.com/) |
+
+The redirects are generated by `scripts/build_redirects.py`, and `tests/`
+verifies the map still covers every URL the old site served. GitHub Pages has
+no server-side redirect table, so each stub is a `rel=canonical` plus a
+`<meta refresh>` — the canonical is what moves search ranking to the new URL.
+
+Two hosts serve this repository, so both are covered: GitHub Pages gets the
+stubs, and `netlify.toml` gives the Netlify project real 301s and stops it
+building the old site from a command configured in its web UI.
+
+## Why the redirects come before the archive
+
+Archiving a repository makes it read-only and stops its Actions from running,
+but it does **not** unpublish its GitHub Pages site: whatever was deployed last
+keeps being served. So the redirects have to be deployed *before* the archive
+switch is flipped. Afterwards there is no way to change what this domain serves
+without unarchiving first.

netlify.toml

@@ -0,0 +1,33 @@
+# This repository is also wired to a Netlify project, which until now built the
+# mkdocs site from a command configured in Netlify's web UI. That command is
+# invisible from here and outlives any change to the GitHub Actions workflow,
+# so without this file Netlify would keep publishing the old site after the
+# GitHub Pages deploy had been switched to redirects — and the deploy preview
+# on a pull request would show the old site too, which makes the redirects look
+# broken when they are not.
+#
+# Settings here override the UI, so the build is defined in version control.
+
+[build]
+  command = "python scripts/build_redirects.py"
+  publish = "site"
+
+# Netlify, unlike GitHub Pages, can serve a real 301. These take precedence
+# over the stub files the build writes, which stay as the fallback for hosts
+# without a redirect table.
+#
+# Order matters: Netlify takes the first rule that matches, so the one page
+# that did not survive as its own URL has to come before the catch-all.
+[[redirects]]
+  from = "/projects/*"
+  to = "https://commit-check.com/"
+  status = 301
+  force = true
+
+# Everything else kept its path on the new site, so the tail is carried across
+# unchanged. This also covers URLs the stub list does not enumerate.
+[[redirects]]
+  from = "/*"
+  to = "https://commit-check.com/:splat"
+  status = 301
+  force = true

scripts/build_redirects.py

@@ -0,0 +1,113 @@
+"""Build the redirect-only site this repository now publishes.
+
+The site moved to https://commit-check.com. This repository is kept as the
+historical source — ``docs/`` is still here and still readable — but what it
+publishes is a set of redirect stubs, one per URL the old site served, so links
+already out in the world land on the page that replaced them.
+
+Why stubs rather than a server redirect: GitHub Pages serves static files and
+has no redirect table, so a ``<meta refresh>`` plus a ``rel=canonical`` is the
+only mechanism available. The canonical link is what transfers search ranking
+to the new URL; the meta refresh and the script are what move a reader.
+
+Deliberately no ``robots: noindex`` on these stubs. It reads like the tidy thing
+to do, but it contradicts the canonical: one says "consolidate this page onto
+that URL", the other says "drop this page from the index", and a crawler that
+honours the second may never act on the first — or carry the noindex across to
+the target. A migration wants the canonical to be believed.
+
+Why this runs instead of ``mkdocs build``: once this repository is archived,
+Actions stop running and the last deployed artifact is what Pages serves
+forever. That artifact needs to be the redirects, so the redirects have to be
+deployed *before* the archive switch is flipped, not after.
+
+Run with ``python scripts/build_redirects.py`` — output goes to ``site/``.
+"""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+NEW_SITE = "https://commit-check.com"
+
+#: Every URL the mkdocs site served, taken from its own build output, mapped to
+#: the path that replaced it on the new site.
+#
+# All but one are the same path: the pages, the blog, its archive, author and
+# category indexes were carried over unchanged, and the posts kept their
+# filenames and ``created`` dates, so the generated slugs match byte for byte.
+#
+# ``/projects/`` is the exception. It was folded into the Ecosystem section of
+# the new landing page, so it redirects to the root rather than to a page that
+# does not exist.
+SAME_PATH = [
+    "/",
+    "/getting-started/",
+    "/blog/",
+    "/blog/2026/06/21/ai-native-json-output-and-a-python-api/",
+    "/blog/2026/06/21/from-zero-config-to-org-wide-policy/",
+    "/blog/2026/06/21/one-policy-file-for-your-git-history/",
+    "/blog/2026/07/06/ai-attribution-governance-enforcing-ai-disclosure-policies-at-the-ci-level/",
+    "/blog/archive/2026/",
+    "/blog/author/team/",
+    "/blog/category/announcements/",
+    "/blog/category/updates/",
+]
+
+REDIRECTS = {path: path for path in SAME_PATH} | {"/projects/": "/"}
+
+# The fragment is carried across by the script: a reader following a deep link
+# into a page should keep their place. ``location.replace`` rather than
+# ``location.href`` so the stub does not land in the back-button history and
+# trap them in a loop between the two sites.
+TEMPLATE = """<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>Moved to commit-check.com</title>
+<link rel="canonical" href="{target}">
+<meta http-equiv="refresh" content="0; url={target}">
+<script>location.replace("{target}" + location.hash);</script>
+<style>
+  body {{ font-family: system-ui, sans-serif; margin: 4rem auto; max-width: 34rem;
+         padding: 0 1rem; line-height: 1.6; }}
+  a {{ color: #2c9ccd; }}
+</style>
+</head>
+<body>
+<h1>This site has moved</h1>
+<p>The Commit Check documentation, landing page and blog are now published at
+<a href="{target}">{target}</a>.</p>
+<p>If you are not redirected automatically, follow the link above.</p>
+</body>
+</html>
+"""
+
+
+def main() -> int:
+    site = Path(__file__).resolve().parent.parent / "site"
+    for old, new in REDIRECTS.items():
+        target = NEW_SITE + new
+        page = site / old.strip("/") / "index.html"
+        page.parent.mkdir(parents=True, exist_ok=True)
+        page.write_text(TEMPLATE.format(target=target), encoding="utf-8")
+
+    # Pages serves this for any path with no file of its own, which covers the
+    # URLs this list missed — a stray deep link, a page from an older layout.
+    # It points at the new site's root because there is nothing better to guess.
+    (site / "404.html").write_text(
+        TEMPLATE.format(target=NEW_SITE + "/"), encoding="utf-8"
+    )
+
+    # Without this, Pages runs the output through Jekyll, which skips files and
+    # directories whose names begin with an underscore.
+    (site / ".nojekyll").write_text("", encoding="utf-8")
+
+    print(f"wrote {len(REDIRECTS)} redirects + 404 fallback to {site}")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

tests/redirects_test.py

@@ -0,0 +1,69 @@
+"""Check the redirect map against the site it is replacing.
+
+This repository publishes redirects now, and once it is archived the artifact
+from the last successful run is what GitHub Pages serves for good — Actions do
+not run on an archived repository, so there is no second chance to fix a URL
+that was left out. That makes the map worth checking while it can still change.
+
+The check builds the mkdocs site the old way and compares the URLs it produces
+against the redirect map, in both directions: a URL the old site served with no
+redirect is a link that will break, and a redirect for a URL the old site never
+served is a sign the map was edited by hand and drifted.
+"""
+
+from __future__ import annotations
+
+import subprocess
+import sys
+import tempfile
+from pathlib import Path
+
+import pytest
+
+ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(ROOT / "scripts"))
+
+from build_redirects import REDIRECTS  # noqa: E402
+
+
+def _urls_the_old_site_served() -> set[str]:
+    """Build the mkdocs site and return every URL it publishes."""
+    with tempfile.TemporaryDirectory() as tmp:
+        result = subprocess.run(
+            [sys.executable, "-m", "mkdocs", "build", "--site-dir", tmp],
+            cwd=ROOT,
+            capture_output=True,
+            text=True,
+        )
+        if result.returncode != 0:
+            # Not a skip. A skip exits pytest 0, the workflow reads that as a
+            # pass and deploys redirects that were never checked against the
+            # site they replace — on the one deploy that cannot be redone.
+            pytest.fail(f"mkdocs build failed: {result.stderr.strip()[:200]}")
+        site = Path(tmp)
+        urls = set()
+        for page in site.rglob("index.html"):
+            rel = page.parent.relative_to(site).as_posix()
+            # ``relative_to`` gives "." for the site root, which is "/".
+            urls.add("/" if rel == "." else f"/{rel}/")
+        return urls
+
+
+def test_every_published_url_has_a_redirect():
+    """Nothing the old site served may be left without a forwarding address."""
+    missing = sorted(_urls_the_old_site_served() - set(REDIRECTS))
+    assert not missing, (
+        "these URLs are served by the current site but have no redirect, so "
+        "they will break when this repository is archived:\n  "
+        + "\n  ".join(missing)
+    )
+
+
+def test_no_redirect_points_at_a_url_that_never_existed():
+    """A redirect for a URL the site never served means the map drifted."""
+    served = _urls_the_old_site_served()
+    invented = sorted(set(REDIRECTS) - served)
+    assert not invented, (
+        "these redirects are for URLs the site does not serve:\n  "
+        + "\n  ".join(invented)
+    )