[saqibmanan]

Source

• Refactor AI model and dataset components to utilize public URLs Slug DataSpaceFrontend#524 "Refactor AI model and dataset components to utilize public URLs Slug" — merged into dev 2026-11-06T06:24:34Z, merge commit e4f3783d9f9ca9236e85986d54a17933bb5f92ff.
• Refactor AI model and dataset components to utilize separate query fi… DataSpaceFrontend#525 "Refactor AI model and dataset components to utilize separate query fi…" — merged into dev 2026-11-06T07:44:43Z, merge commit 8c4d7c62a7d555870677c9c4e76a638fdc2e7e9e.

Both on the same provider-flow-update branch, ~80 minutes apart. #525 is a same-day refactor of #524's own feature (moves queries into queries.ts files, deletes lib/canonical-redirect.ts, simplifies the redirect condition to fire only for a legacy integer/UUID id). Treated as one coverage unit: what matters is the behaviour that survived into the final, deployed state, not either diff in isolation.

Confirmed deployed: DataSpaceFrontend "Deploy DataSpace" run for the #525 merge commit (run 42182886658) completed with promote → success and rollback → skipped. Live on dev now.

What actually changed (from the diff, not the titles)

lib/public-link.ts (new) introduces isUuid, isIntegerId, publicHref, publicPageUrl, datasetLookup, aiModelLookup. Applied to:

• Canonical redirects: /datasets/[datasetIdentifier], /aimodels/[modelId], /usecases/[useCaseSlug] now 308-redirect a legacy raw id (dataset UUID, AI model/use case integer id) to /<section>/<urlSlug>. Visiting the urlSlug address directly serves with no further redirect.
• Listing/card links: home Datasets/UseCases, ListingComponent, publishers' Datasets/UseCases, search's UnifiedListingComponent, dataset SimilarDatasets/DatasetConnected, usecase DatasetStories, the AI model/use case publish-preview links, and the collaboratives detail page all switched from linking by raw id to urlSlug (falling back to id).
• Sitemap route (app/sitemap/[entityPage]/route.ts): dataset/aimodel/usecase <url><loc> entries switched from raw id to urlSlug.
• A correctness fix bundled into the same unit: DatasetDetailsPage's chart-count query and analytics trackDataset call now wait for the dataset's real UUID (recordId, from the GraphQL response) instead of firing on whatever string is in the URL — needed because that URL segment can now be a slug, not a UUID.

Categories added, and why

Per the test-sync policy's §3 (new/changed UI page behaviour → smoke, +functional), and since the whole user-visible surface here is "does the right page load at the right address" — a smoke-level concern. Added to tests/api/smoke/test_api_053_canonical_url_slug_redirect.py:

• test_dataset_legacy_id_redirects_to_canonical_slug
• test_aimodel_legacy_id_redirects_to_canonical_slug
• test_usecase_legacy_id_redirects_to_canonical_slug
• test_dataset_canonical_slug_serves_without_redirecting
• test_aimodel_canonical_slug_serves_without_redirecting
• test_usecase_canonical_slug_serves_without_redirecting

All unauthenticated (public detail pages, no login) — this run has no test credentials, so coverage is scoped to what can be proven without them.

Not readonly: git merge-base --is-ancestor 8c4d7c62a7d555870677c9c4e76a638fdc2e7e9e origin/main on DataSpaceFrontend is false — this hasn't shipped to main/prod. Prod's backend GraphQL schema doesn't even have a urlSlug field yet (Cannot query field 'urlSlug' on type 'TypeDataset', confirmed live against api.datakeep.civicdays.in). Marking this readonly would fail every DataSpaceBackend/DataSpaceFrontend main deploy's smoke gate until the backend half ships too. Gated with deployed_pr("DataSpaceFrontend#525", site="HOME_URL_DEV") instead — skips wherever the frontend under test doesn't yet serve the merge commit, and turns on by itself once it does. Follow-up: add readonly once #525 is on main.

Red → green proof

GREEN, today, on dev (HOME_URL_DEV=https://dev.civicdataspace.in API_BASE_URL=https://dev.api.civicdataspace.in):

tests/api/smoke/test_api_053_canonical_url_slug_redirect.py::test_dataset_legacy_id_redirects_to_canonical_slug PASSED
tests/api/smoke/test_api_053_canonical_url_slug_redirect.py::test_aimodel_legacy_id_redirects_to_canonical_slug PASSED
tests/api/smoke/test_api_053_canonical_url_slug_redirect.py::test_usecase_legacy_id_redirects_to_canonical_slug PASSED
tests/api/smoke/test_api_053_canonical_url_slug_redirect.py::test_dataset_canonical_slug_serves_without_redirecting PASSED
tests/api/smoke/test_api_053_canonical_url_slug_redirect.py::test_aimodel_canonical_slug_serves_without_redirecting PASSED
tests/api/smoke/test_api_053_canonical_url_slug_redirect.py::test_usecase_canonical_slug_serves_without_redirecting PASSED
6 passed in 7.69s

RED, today, on prod (HOME_URL_DEV=https://civicdataspace.in API_BASE_URL=https://api.datakeep.civicdays.in), which still runs the pre-#524 code — a real "parent commit" red, not a flipped assertion. Ran with the deployed_pr marker temporarily stripped locally (restored before committing — final diff has no test-content changes, only the marker gate):

FAILED test_dataset_legacy_id_redirects_to_canonical_slug
  AssertionError: expected a 308 permanent redirect off the raw dataset id 'd4134c28-1b40-4e16-a3c0-cecd672e9cfc', got 200.
FAILED test_aimodel_legacy_id_redirects_to_canonical_slug
  AssertionError: expected a 308 permanent redirect off the raw AI model id 46, got 200.
FAILED test_usecase_legacy_id_redirects_to_canonical_slug
  AssertionError: expected a 308 permanent redirect off the raw use case id '314', got 200.
3 failed in 3.49s

(Only the 3 legacy-id tests are schema-safe to point at prod — the 3 canonical-slug tests query urlSlug in their fixtures, which doesn't exist on prod's backend schema, so they're dev-only by construction and weren't part of the red run.)

With the deployed_pr marker restored and pointed at prod, the 3 legacy-id tests correctly skip instead of failing: deployed_pr DataSpaceFrontend#525: https://civicdataspace.in runs 92cd5219, which doesn't include 8c4d7c62 yet — confirming the gate itself works.

Collection check, matching run-smoke.yml's real pull_request filter (tests/api/smoke -m smoke):

$ pytest tests/api/smoke -m "smoke" --collect-only -q | grep 053
<all 6 tests listed>

Gaps — deliberately not covered here

• Listing/card links now use urlSlug (home, search, publishers, SimilarDatasets, DatasetConnected, DatasetStories, collaboratives) — this is React-client-rendered (confirmed: curl'ing /datasets returns no server-rendered href in the initial HTML), so checking it needs a real browser, not an HTTP client. This cloud run has no credentials but these are public pages, so a browser check wouldn't need any — it's simply out of scope for an HTTP-only unauthenticated pass. A local/authenticated run should add a Selenium check (existing locators//pages/ Page Objects for dataset/use case listing cards) asserting card hrefs end in the item's urlSlug, not its raw id.
• Sitemap route's switch to urlSlug (app/sitemap/[entityPage]/route.ts) — untestable right now: dev's sitemap is disabled by design (DataSpaceFrontend#493, covered by test_api_027_dev_noindex.py; confirmed live: /sitemap.xml on dev returns "Sitemaps are not enabled"), and the feature isn't on prod yet either. Once #525 (or whichever PR carries it) reaches main, extend test_api_006_seo_sitemap.py's prod checks to assert dataset/aimodel/usecase <url><loc> entries end in the backend's urlSlug, cross-checked per record, not just that the count matches.
• DatasetDetailsPage's chart-count/analytics fix (waiting for recordId before firing) — a real bug fix bundled into this diff, but it only matters once a dataset is reached by its slug (which the canonical-redirect tests above already exercise end-to-end as part of a normal page load); no separate assertion added, since the existing chart-count and analytics calls aren't independently observable without a browser and TanStack Query internals.
• No regression marker used: regression is never selected by any CI event here (deselected on push/PR/dispatch alike), so a regression-only test would merge and silently never run. Everything above carries smoke + api instead.

Safe to merge

Yes — not gated on the backend shipping anything further; deployed_pr handles the only real hazard (dev rolling back or not yet carrying #525), and this file is not readonly so it never touches the main-deploy prod gate.