feat(tests/playwright): rendering invariant coverage — 10 new tests

[avrabe]

Summary

Audit-driven Playwright coverage expansion. Cross-referenced every .route(...) declaration in rivet-cli/src/serve/mod.rs against every page.goto(...) / page.request.get(...) in tests/playwright/*.spec.ts to identify rendering paths the dashboard silently relies on but that no test exercises end-to-end.

10 new tests pinned in tests/playwright/rendering-invariants.spec.ts. No production code was changed — all audit findings are either already-correct behavior worth pinning, or known asymmetries that should be gated against silent regression.

Audit table

Route	Status before	Test added
/	Covered	—
/artifacts	Deeply covered	—
/artifacts/{id}	Covered	— for ? happy path; NEW for unknown ID
/artifacts/{id}/preview	None	NEW
/artifacts/{id}/graph	None	NEW
/validate	Covered	—
/matrix	Covered	—
/matrix/cell	None	NEW
/graph	Covered	NEW for ?variant=...
/stats	Covered	—
/coverage	Covered	—
/documents, /documents/{id}	Covered	—
/search	Smoke (status only)	NEW for empty + no-results fragments
/verification	Covered	—
/stpa	Deeply covered	—
/eu-ai-act	None	NEW
/results, /results/{run_id}	Covered	—
/source, /source/{*path}	Covered	—
/diff	Covered	—
/doc-linkage	Covered	—
/traceability, /traceability/history	Covered	—
/help, /help/schema, /help/links, /help/rules, /help/docs	Covered	—
/help/docs/{*slug}	None	NEW
/externals, /externals/{prefix}	Covered	—
/variants	Covered	—
/oembed	Covered (api.spec)	—
/embed/*	None (URL pattern referenced by oembed but never navigated)	NEW
/api/v1/*	Covered (api.spec)	—
/assets/*, /wasm/*, /source-raw/*, /docs-asset/*	Covered (asset-loading.spec)	—

Per-test summary

1. /eu-ai-act — full-page render through middleware. Pins that the route renders the EU-AI-Act compliance heading and the layout middleware wraps it with <nav>. Whole route was a black box before.

2. /help/docs/{slug} — topic page renders with back-link. /help/docs index was tested; clicking through to a topic was not. Pins the <a href=\"/help/docs\"> back-link contract on a real slug (cli).

3. /artifacts/{id}/preview — hover-tooltip fragment shape. This endpoint is hit by hx-get on hover but never directly navigated. Pins the .art-preview / .art-preview-header / .art-preview-title class hierarchy that the hover-tooltip CSS relies on.

4. /artifacts/{id}/graph — ego-graph wraps in #ego-graph-viewer. The diagram-viewer parity test (diagram-viewer.spec.ts) covered /graph, /doc-linkage, /help/schema but not the per-artifact ego-graph. Closes that gap with the same toolbar contract.

5. /matrix/cell — <ul> fragment as direct child of main#content. The matrix-cell drill-down (HTMX-loaded into /matrix) was never exercised end-to-end. Pins the wrapper shape so the matrix view's HTMX target keeps working.

6. /embed/{url} — embed_layout (no nav, no .shell). The oEmbed test in api.spec.ts confirms the URL pattern is generated but never navigates to it. Pins the structural difference: embed has <main id=\"content\"> but no <nav role=\"navigation\"> and no body > .shell — the contract VS Code WebView relies on.

7. Mermaid in artifact description — emitted as <pre class=\"mermaid\">, NOT wrapped in .svg-viewer. ARCH-CORE-001 has a fenced ```mermaid block in its description (see artifacts/architecture.yaml). Pins TWO things: (a) the markdown renderer recognises and emits <pre.mermaid>, (b) description-mermaid is currently NOT wrapped in .svg-viewer (only the dedicated diagram: field is). This is a known UX asymmetry; pinning it forces any future fix to be intentional.

8. /graph?variant=... silently unscoped. graph_view uses GraphParams (not ViewParams) and has no variant field — the URL param is dropped at the handler level. But the layout middleware extracts variant= from the URL and injects the variant-banner anyway. Pins the surprising current state: banner says "Filtered to variant: minimal-ci" but the graph is the unfiltered full graph. Gated against silent change.

9. /artifacts/UNKNOWN-ID returns 200 (not 404) with "Not Found" body. artifact_detail always returns Html(...).into_response() — i.e. status 200 — even when the artifact is missing. Consistent with /externals/<unknown-prefix> (already pinned at externals.spec.ts:80) but inconsistent with the gut expectation. Now gated.

10. /search empty-query + no-results fragments emit .cmd-k-empty. routes.spec.ts smoke-tested /search?q=OSLC. Both edge cases (no query, no results) emit a specific .cmd-k-empty placeholder with distinct copy ("Type to search" vs "No results"). The cmd-k overlay UI keys off this class.

Audit gaps NOT pinned (deferred)

The audit also surfaced rendering invariants that warrant attention but weren't pinned in this PR — they require a product decision rather than a test:

• Variant-scoping coverage gap. Beyond graph_view, these handlers also lack a variant param: verification_view, eu_ai_act_view, traceability_view, doc_linkage_view, documents_list, results_view, externals_list, search_view. The layout still shows the banner for any URL with ?variant=.... Test feat: cross-repository artifact linking #8 pins the most prominent case (/graph); the others are similarly inconsistent and should probably either all opt in or all opt out. Tracked separately.
• Empty-state structural pins. /results (no test results), /externals (no externals configured), /coverage (empty rules) all have bespoke empty-state HTML that's never exercised in dogfood data. These need a feature-flagged test fixture rather than a runtime test.
• HTMX-vs-direct rendering parity. The wrap_full_page middleware decides per-request whether to wrap in layout. The decision tree (is_htmx, is_print, is_embed, path prefixes) is non-trivial; only the obvious cases are tested.

Test plan

• npx playwright test tests/playwright/rendering-invariants.spec.ts passes locally
• CI runs the full Playwright suite — no regression in existing specs
• If any test fails on CI, the failure points to either a real regression or a deliberate behavior change that the test correctly gated

Refs: FEAT-001

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Rivet Criterion Benchmarks'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: bdc9a10	Previous: 58801b6	Ratio
store_lookup/1000	24942 ns/iter (± 38)	19320 ns/iter (± 203)	1.29
traceability_matrix/100	4918 ns/iter (± 51)	3903 ns/iter (± 7)	1.26
diff/1000	719936 ns/iter (± 14484)	597822 ns/iter (± 2900)	1.20
query/100	773 ns/iter (± 7)	630 ns/iter (± 1)	1.23
query/1000	6799 ns/iter (± 183)	5229 ns/iter (± 124)	1.30

This comment was automatically generated by workflow using github-action-benchmark.