fix(deploy): stop hydrating content/data from empty publish branch (#627)

[jmservera]

Fixes #627.

Root cause

PR #623 added content/data/ to the deploy "Hydrate generated content from publish" step in .github/workflows/deploy-site.yml. The publish branch (last updated 2026-07-27, pre-#623) does not carry the observatory data pages, so the step's rm -rf "$path" followed by a failing git checkout origin/publish -- "$path" 2>/dev/null || true left content/data/ empty at deploy time.

The committed embed page content/embeds/fastest-growing-ai-repositories-chart/ (not in the hydration list) then referenced the now-missing /data/fastest-growing-ai-repositories-this-year/, and the errorf guards in layouts/embeds/single.html and layouts/shortcodes/observatory-chart.html aborted the build:

ERROR observatory-chart shortcode could not find page "/data/fastest-growing-ai-repositories-this-year/"
ERROR embed page embeds/fastest-growing-ai-repositories-chart/index.md is missing a valid source_page

Why CI didn't catch it

CI's Site Preview and Production site jobs build the repository content as committed on the branch, where content/data/ exists and the embed resolves. Only deploy-site.yml runs the publish-hydration step, so the main (code + embeds) ↔ publish (generated data) divergence is structurally invisible to CI.

Fix (Path B — interim unblock)

Omit content/data/ from the deploy hydration list so the committed pages ship until the crawl repopulates publish. content/data/ is still carried through the crawl-and-publish transaction; the provenance-invariant test is updated to reflect the intentional deploy exclusion.

Validation

• pytest tests/test_pipeline.py — 28 passed
• zizmor .github/workflows/deploy-site.yml — no findings
• hugo --minify — clean build (exit 0); the embed and its data page render
• ruff check . — clean

Follow-up (RPI handoff)

• PRD updated: NFR-011 (deploy/CI build parity), NFR-012 (embed source_page must resolve), R-08 (hydration wipes committed pages), Q-03 (when to restore hydration), changelog v1.2.
• New plan: .copilot-tracking/plans/2026-07-31/claracle-deploy-hydration-remediation-plan.instructions.md covering:
  1. Interim unblock (this PR)
  2. Regenerate observatory content onto publish (run the crawl/generation actions)
  3. Restore consistent deploy hydration
  4. Add a CI deploy-parity guard
  5. Validation + close 🔴 Deploy Hugo site failed (run 30616770243) #627

Note: this ships the committed (seed) content/data pages to production until the crawl publishes real observatory data. Restoring hydration (Phase 3) depends on running the generation actions first (Phase 2).

[Copilot]

Pull request overview

This PR unblocks GitHub Pages deploys by preventing the deploy workflow from deleting committed Observatory data pages when hydrating from an out-of-date publish branch, and aligns the pipeline provenance test + documentation with that interim exception (issue #627).

Changes:

• Stop deploy-site.yml from hydrating content/data/ from publish (temporary exception to avoid wiping required pages).
• Update tests/test_pipeline.py provenance invariant to exclude content/data/ from deploy hydration while still requiring it in the publish transaction.
• Record the incident + interim mitigation in the PRD and add a remediation plan document for follow-up work.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

File	Description
.github/workflows/deploy-site.yml	Removes content/data/ from deploy hydration list to prevent build-breaking deletions when publish is missing those pages.
tests/test_pipeline.py	Adjusts pipeline invariant expectations so deploy hydration intentionally excludes content/data/.
docs/prds/claracle-data-observatory-relaunch.md	Documents deploy/CI parity requirements and records the incident/fix in v1.2.
.copilot-tracking/plans/2026-07-31/claracle-deploy-hydration-remediation-plan.instructions.md	Adds a structured follow-up plan for regenerating publish, restoring hydration, and adding CI parity guards.

Suppressed comments (1)

.github/workflows/deploy-site.yml:99

• The new explanatory comment doesn’t name the omitted path, which makes it harder to understand what needs to be re-added later. Calling out content/data/ explicitly will reduce the chance of reintroducing issue #627.

          # The observatory data-page directory is intentionally omitted here:
          # those pages are not yet present on publish, so hydrating an empty
          # source would delete the committed pages that the chart embeds depend
          # on and break the build (see issue #627). Re-add it once the crawl
          # publishes those pages.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.

Suppressed comments (1)

tests/test_pipeline.py:652

• deploy_hydrate_entries check only asserts the exact line content/data/ is absent. If someone later adds content/data (without the trailing slash) to GENERATED_PATHS, deploy would still delete the directory (because the script normalizes with ${path%/}), but this test would not catch the regression. Normalize trailing slashes (or assert both variants) before checking membership.

        deploy_hydrate_entries = {line.strip() for line in deploy_hydrate.splitlines()}
        self.assertNotIn("content/data/", deploy_hydrate_entries)