feat: blog auto-publish cron

.github/workflows/blog-autopublish.yml

@@ -0,0 +1,178 @@
+name: Blog auto-publish
+
+# Runs daily at 06:00 UTC. Scans content/blog/ for drafts; for any
+# post with `date <= today`, `draft = true`, and `hold` not set, it
+# opens an auto-merge PR that flips draft to false. Each publish
+# becomes its own PR/squash-merge so deploy fires per-post and the
+# audit trail is clean.
+#
+# Always posts a status report to a pinned issue (label
+# `autopublish-status`) using a bot-marker (`<!-- blog-cron-status -->`)
+# so the same comment is replaced each run instead of appended.
+#
+# Failure path opens a separate issue tagged `autopublish-failure`.
+#
+# Manual trigger via workflow_dispatch is supported for first-run /
+# debugging. Concurrency is serialized so two runs cannot interleave
+# (e.g. cron + manual).
+#
+# Security note: this workflow has only `schedule` and
+# `workflow_dispatch` triggers — neither carries untrusted user input.
+# All values flowing into `run:` blocks come from server-controlled
+# context (github.run_id, github.repository) or from our own script
+# reading our own repository's frontmatter, so no command-injection
+# surface exists from external actors.
+
+on:
+  schedule:
+    - cron: "0 6 * * *"
+  workflow_dispatch:
+
+permissions:
+  contents: write
+  pull-requests: write
+  issues: write
+
+concurrency:
+  group: blog-autopublish
+  cancel-in-progress: false
+
+jobs:
+  publish:
+    name: Scan, publish, report
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+    env:
+      RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+      REPO: ${{ github.repository }}
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Configure git identity
+        run: |
+          git config user.name "github-actions[bot]"
+          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+
+      - name: Scan drafts
+        id: scan
+        run: |
+          set -euo pipefail
+          python3 scripts/blog-autopublish.py --mode scan > /tmp/inventory.json
+          ready_count=$(jq '.ready | length' /tmp/inventory.json)
+          scheduled_count=$(jq '.scheduled | length' /tmp/inventory.json)
+          held_count=$(jq '.held | length' /tmp/inventory.json)
+          echo "ready_count=${ready_count}" >> "$GITHUB_OUTPUT"
+          echo "scheduled_count=${scheduled_count}" >> "$GITHUB_OUTPUT"
+          echo "held_count=${held_count}" >> "$GITHUB_OUTPUT"
+          {
+            echo "## Inventory"
+            echo '```json'
+            cat /tmp/inventory.json
+            echo '```'
+          } >> "$GITHUB_STEP_SUMMARY"
+
+      - name: Publish ready posts
+        id: publish
+        if: steps.scan.outputs.ready_count != '0'
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          set -euo pipefail
+          published='[]'
+          today=$(date -u +%Y-%m-%d)
+
+          # Iterate each ready post, one PR per post.
+          while IFS=$'\t' read -r path slug; do
+            branch="cron/publish-${slug}-${today}"
+
+            # Clear any stale branch from a previous failed run.
+            git branch -D "$branch" 2>/dev/null || true
+            git push origin --delete "$branch" 2>/dev/null || true
+
+            git checkout -b "$branch"
+            python3 scripts/blog-autopublish.py --mode flip --file "$path"
+            git add "$path"
+            git commit -m "publish: ${slug}"
+            git push -u origin "$branch"
+
+            pr_url=$(gh pr create \
+              --base main \
+              --head "$branch" \
+              --title "publish: ${slug}" \
+              --body "Automated publish for \`${slug}\` triggered by date \`${today}\`. Opened by the blog-autopublish workflow." \
+              --label autopublish)
+            pr_num=$(echo "$pr_url" | grep -oE '[0-9]+$')
+
+            # Squash-merge directly. If branch protections require checks,
+            # fall back to --auto so the merge happens once CI passes.
+            gh pr merge "$pr_num" --squash --delete-branch || \
+              gh pr merge "$pr_num" --auto --squash --delete-branch
+
+            published=$(echo "$published" | jq --arg s "$slug" --arg p "$pr_num" '. + [{slug: $s, pr: $p}]')
+            git checkout main
+            git pull --ff-only origin main
+          done < <(jq -r '.ready[] | [.path, .slug] | @tsv' /tmp/inventory.json)
+
+          echo "$published" > /tmp/published.json
+
+      - name: Render status report
+        if: always()
+        run: |
+          set -euo pipefail
+          published=$(cat /tmp/published.json 2>/dev/null || echo '[]')
+          python3 scripts/blog-autopublish.py \
+            --mode report \
+            --inventory /tmp/inventory.json \
+            --published "$published" \
+            --run-url "$RUN_URL" > /tmp/report.md
+          cat /tmp/report.md >> "$GITHUB_STEP_SUMMARY"
+
+      - name: Post / update status comment
+        if: always()
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          set -euo pipefail
+
+          # Locate the pinned status issue by label, or create it on first run.
+          issue_num=$(gh issue list --label autopublish-status --state open --json number --jq '.[0].number // empty')
+          if [ -z "$issue_num" ]; then
+            issue_url=$(gh issue create \
+              --title "Blog auto-publish status" \
+              --label autopublish-status \
+              --body-file /tmp/report.md)
+            issue_num=$(echo "$issue_url" | grep -oE '[0-9]+$')
+            echo "Created status issue #${issue_num}"
+            exit 0
+          fi
+
+          # Find the previous bot comment by its hidden marker.
+          comment_id=$(gh api "repos/${REPO}/issues/${issue_num}/comments" --paginate \
+            --jq '.[] | select(.body | contains("<!-- blog-cron-status -->")) | .id' | head -1)
+
+          if [ -n "$comment_id" ]; then
+            gh api "repos/${REPO}/issues/comments/${comment_id}" \
+              -X PATCH \
+              --field body=@/tmp/report.md
+            echo "Updated comment ${comment_id} on issue #${issue_num}"
+          else
+            gh issue comment "$issue_num" --body-file /tmp/report.md
+            echo "Created new comment on issue #${issue_num}"
+          fi
+
+      - name: Open failure issue
+        if: failure()
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          today=$(date -u +%Y-%m-%d)
+          gh issue create \
+            --title "blog-autopublish failed: ${today}" \
+            --label autopublish-failure \
+            --body "The daily blog-autopublish workflow failed on ${today}. Logs: ${RUN_URL}"

scripts/blog-autopublish.py

@@ -0,0 +1,167 @@
+#!/usr/bin/env python3
+"""Blog auto-publish: scan content/blog/ drafts and categorize them by schedule.
+
+Three modes:
+
+  --mode scan
+      Print a JSON inventory to stdout: {today, ready, scheduled, held}.
+      A draft is "ready" when date <= today and hold is not true.
+      A draft is "scheduled" when date > today (and hold is not true).
+      A draft is "held" when hold = true (regardless of date).
+
+  --mode flip --file PATH
+      Atomically replace the first `draft = true` line in PATH with
+      `draft = false`. Used by the workflow once a post moves to ready.
+
+  --mode report --inventory PATH --published JSON [--run-url URL]
+      Render a markdown status report (with a `<!-- blog-cron-status -->`
+      bot marker) suitable for posting on the pinned status issue.
+
+The frontmatter parser is intentionally regex-based, not full-TOML — it
+only reads the four fields that drive the cron (date, draft, hold, title)
+and does not need to interpret arrays or tables. This avoids a dependency
+on `tomllib` and keeps the script readable.
+"""
+
+import argparse
+import datetime
+import json
+import pathlib
+import re
+import sys
+
+BLOG_DIR = pathlib.Path("content/blog")
+
+FRONTMATTER_RE = re.compile(r"\A\+\+\+\n(.*?)\n\+\+\+", re.S)
+DATE_RE = re.compile(r"^date\s*=\s*(\d{4}-\d{2}-\d{2})\s*$", re.M)
+DRAFT_RE = re.compile(r"^draft\s*=\s*(true|false)\s*$", re.M)
+HOLD_RE = re.compile(r"^hold\s*=\s*(true|false)\s*$", re.M)
+TITLE_RE = re.compile(r'^title\s*=\s*"([^"]*)"\s*$', re.M)
+DRAFT_TRUE_LINE = re.compile(r"^draft\s*=\s*true\s*$", re.M)
+
+
+def parse_post(path: pathlib.Path) -> dict | None:
+    """Return a record for a post, or None if it has no parseable frontmatter."""
+    text = path.read_text(encoding="utf-8")
+    fm_match = FRONTMATTER_RE.match(text)
+    if not fm_match:
+        return None
+    fm = fm_match.group(1)
+    date_m = DATE_RE.search(fm)
+    draft_m = DRAFT_RE.search(fm)
+    hold_m = HOLD_RE.search(fm)
+    title_m = TITLE_RE.search(fm)
+    return {
+        "path": str(path),
+        "slug": path.stem.split("-", 3)[-1] if path.stem[:10].count("-") == 2 else path.stem,
+        "title": title_m.group(1) if title_m else path.stem,
+        "date": date_m.group(1) if date_m else None,
+        "draft": bool(draft_m and draft_m.group(1) == "true"),
+        "hold": bool(hold_m and hold_m.group(1) == "true"),
+    }
+
+
+def scan(today: str) -> dict:
+    drafts = []
+    for path in sorted(BLOG_DIR.glob("*.md")):
+        if path.name.startswith("_"):
+            continue
+        record = parse_post(path)
+        if record is None:
+            continue
+        if record["draft"]:
+            drafts.append(record)
+    held = [p for p in drafts if p["hold"]]
+    active = [p for p in drafts if not p["hold"]]
+    ready = [p for p in active if p["date"] and p["date"] <= today]
+    scheduled = [p for p in active if p["date"] and p["date"] > today]
+    return {
+        "today": today,
+        "ready": ready,
+        "scheduled": sorted(scheduled, key=lambda p: p["date"]),
+        "held": sorted(held, key=lambda p: p["slug"]),
+    }
+
+
+def flip(path_str: str) -> None:
+    path = pathlib.Path(path_str)
+    text = path.read_text(encoding="utf-8")
+    new_text, count = DRAFT_TRUE_LINE.subn("draft = false", text, count=1)
+    if count == 0:
+        sys.exit(f"flip: no `draft = true` line found in {path_str}")
+    path.write_text(new_text, encoding="utf-8")
+
+
+def render_report(inventory: dict, published: list[dict], run_url: str) -> str:
+    today = inventory["today"]
+    lines = [
+        "<!-- blog-cron-status -->",
+        f"## Blog auto-publish · last run {today}",
+        "",
+    ]
+    if published:
+        lines.append(f"**This run:** published {len(published)} —")
+        for entry in published:
+            slug = entry["slug"]
+            pr = entry.get("pr")
+            pr_suffix = f" (PR #{pr})" if pr else ""
+            lines.append(f"- `{slug}`{pr_suffix}")
+    else:
+        lines.append("**This run:** no posts ready to publish today.")
+    lines.append("")
+
+    scheduled = inventory["scheduled"]
+    lines.append(f"**Scheduled ({len(scheduled)}):**")
+    if scheduled:
+        for p in scheduled:
+            lines.append(f"- {p['date']} · `{p['slug']}` — {p['title']}")
+    else:
+        lines.append("- _none_")
+    lines.append("")
+
+    held = inventory["held"]
+    lines.append(f"**Held ({len(held)}):**")
+    if held:
+        for p in held:
+            lines.append(f"- ∞ · `{p['slug']}` — {p['title']}")
+    else:
+        lines.append("- _none_")
+    lines.append("")
+
+    if run_url:
+        lines.append(f"_Run: {run_url}_")
+    return "\n".join(lines) + "\n"
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--mode", choices=["scan", "flip", "report"], required=True)
+    parser.add_argument("--file", help="Path to the post (--mode flip)")
+    parser.add_argument("--inventory", help="Path to inventory JSON (--mode report)")
+    parser.add_argument(
+        "--published",
+        default="[]",
+        help="JSON array of {slug, pr} for posts published this run (--mode report)",
+    )
+    parser.add_argument("--run-url", default="", help="GitHub Actions run URL (--mode report)")
+    args = parser.parse_args()
+
+    if args.mode == "scan":
+        today = datetime.date.today().isoformat()
+        json.dump(scan(today), sys.stdout, indent=2)
+        sys.stdout.write("\n")
+    elif args.mode == "flip":
+        if not args.file:
+            sys.exit("--file is required for --mode flip")
+        flip(args.file)
+    elif args.mode == "report":
+        if not args.inventory:
+            sys.exit("--inventory is required for --mode report")
+        with open(args.inventory, encoding="utf-8") as f:
+            inventory = json.load(f)
+        published = json.loads(args.published)
+        sys.stdout.write(render_report(inventory, published, args.run_url))
+
+
+if __name__ == "__main__":
+    main()