fix(docs): retain readable tabs without JavaScript and verify current staging

Author: emir-karabeg

.github/workflows/docs-negotiation.yml

@@ -159,6 +159,7 @@ jobs:
       DOCS_HEADER_REPORT_PATH: ${{ github.workspace }}/apps/docs/test-results/docs-headers.json
       DOCS_NAVIGATION_REPORT_PATH: ${{ github.workspace }}/apps/docs/test-results/navigation/docs.json
       DOCS_GENERATORS_REPORT_PATH: ${{ github.workspace }}/apps/docs/test-results/generators/report.json
+      DOCS_MIGRATION_REPORT_PATH: ${{ github.workspace }}/apps/docs/test-results/migration.json
       DOCS_SAMPLES_REPORT_PATH: ${{ github.workspace }}/apps/docs/test-results/samples/report.json
       DOCS_SAMPLES_SYNTAX_LANGUAGES: javascript,python,bash,go,java,csharp
       DOCS_SAMPLES_HTTP_BASE_URL: http://127.0.0.1:3001
@@ -180,27 +181,37 @@ jobs:
         run: bun run --cwd apps/docs build
       - name: Verify generator idempotence and authored regions
         run: bun run --cwd apps/docs scripts/test-generators-e2e.ts
+      - name: Verify preserved routes, anchors and examples
+        if: always() && steps.build.outcome == 'success'
+        run: bun run --cwd apps/docs content:verify-migration
       - name: Verify native header edge cases
+        if: always() && steps.build.outcome == 'success'
         run: bun run --cwd apps/docs test:headers:e2e
       - name: Verify isolated PostgreSQL search and index lifecycle
+        if: always() && steps.build.outcome == 'success'
         run: bun run --cwd apps/docs test:search:postgres:e2e
       - name: Set up Java for copied request examples
+        if: always() && steps.build.outcome == 'success'
         uses: actions/setup-java@b6effb05e454b25005698d916606bdc6ffcbf961 # v5
         with:
           distribution: temurin
           java-version: '17'
       - name: Set up .NET for copied request examples
+        if: always() && steps.build.outcome == 'success'
         uses: actions/setup-dotnet@26b0ec14cb23fa6904739307f278c14f94c95bf1 # v5
         with:
           dotnet-version: '8.0.x'
       - name: Set up Go for copied request examples
+        if: always() && steps.build.outcome == 'success'
         uses: actions/setup-go@924ae3a1cded613372ab5595356fb5720e22ba16 # v6
         with:
           go-version: '1.25.x'
           cache: false
       - name: Install browser for navigation checks
+        if: always() && steps.build.outcome == 'success'
         run: bunx --no-install playwright install --with-deps chromium webkit
       - name: Verify HTTP content negotiation
+        if: always() && steps.build.outcome == 'success'
         run: |
           mkdir -p apps/docs/test-results
           bun run --cwd apps/docs start --port 3001 >apps/docs/test-results/docs-server.log 2>&1 &

apps/docs/MACHINE_ACCESS.md

@@ -79,7 +79,7 @@ The PostgreSQL suite uses a new, empty, loopback database and no embedding key.
 
 The platform consumer harness also uses an empty loopback database and the running docs build. From `apps/sim`, run `bun --no-env-file scripts/test-docs-search-provider-e2e.ts` with `DOCS_PROVIDER_E2E_DATABASE_URL`, `DOCS_PROVIDER_E2E_BASE_URL` and `DOCS_PROVIDER_E2E_REPORT_PATH`. Its default mode uses the existing Azure endpoint configuration with a local HTTP fixture, deterministic vectors and the real PostgreSQL provider-admission store. It verifies the platform's direct-table search, scope exclusions, VFS reads and public citations, with no paid requests. This verifies the transport and consumer contract, not semantic relevance.
 
-Live provider verification is opt-in: set `DOCS_PROVIDER_E2E_ALLOW_LIVE=1` and supply `OPENAI_API_KEY`, using a fresh database named `docs_search_provider_test`. The harness limits publication to one introduction chunk and the consumer to one query, with at most two provider requests. Provider failure stops the run and remains a failed report. The default fixture mode also accepts the empty `docs_search_test` database used by the PostgreSQL suite. Both modes remove their tables when finished.
+Live provider verification is opt-in: set `DOCS_PROVIDER_E2E_ALLOW_LIVE=1` and supply `OPENAI_API_KEY`, using a fresh database named `docs_search_provider_test`. The harness limits publication to the introduction page, rejects plans above 15 chunks, and sends at most one batch with up to three attempts. The consumer makes at most one query embedding request, for a total cap of four provider requests. Provider failure stops the run and remains a failed report. The default fixture mode also accepts the empty `docs_search_test` database used by the PostgreSQL suite. Both modes remove their tables when finished.
 
 To repeat locally, create a disposable container (the command returns its ID):
 

apps/docs/components/docs/sample-panels.tsx

@@ -85,6 +85,7 @@ export function SamplePanels({
         <Tab
           key={item.value}
           value={item.value}
+          label={item.label}
           id={item.id ?? `${selectorId}-${item.value}`}
           {...(presentation === 'select' ? { role: 'region', 'aria-labelledby': selectorId } : {})}
           className='[&>div]:pt-0'

apps/docs/components/ui/tabs.tsx

@@ -38,6 +38,7 @@ interface TabsProps extends Omit<ComponentPropsWithoutRef<typeof RadixTabs.Root>
 
 interface TabProps extends Omit<ComponentPropsWithoutRef<typeof RadixTabs.Content>, 'value'> {
   value?: string
+  label?: string
 }
 
 /** Accessible provider tabs retain every panel in HTML for search and deep links. */
@@ -167,6 +168,14 @@ export function Tabs({
         {...props}
         data-docs-tabs-root=''
       >
+        <noscript>
+          <style>{`
+            .docs-root [data-docs-tab-panel][hidden] { display: block; }
+            .docs-root [data-docs-tabs-root] [role="tablist"],
+            .docs-root [data-docs-tabs-root] select,
+            .docs-root [data-docs-tabs-root] button { display: none; }
+          `}</style>
+        </noscript>
         {items && (
           <TabsList aria-label={label}>
             {items.map((item) => (
@@ -243,6 +252,7 @@ export function TabsTrigger({
 
 export function Tab({
   value = '',
+  label,
   id,
   className,
   children,
@@ -269,6 +279,9 @@ export function Tab({
         className
       )}
     >
+      <noscript>
+        <p className='px-3 pt-4 font-medium text-docs-caption'>{label ?? value}</p>
+      </noscript>
       <div className='pt-4 [&>:first-child]:mt-0 [&>:last-child]:mb-0'>{children}</div>
     </RadixTabs.Content>
   )

apps/docs/migration/VERIFICATION.md

@@ -6,6 +6,12 @@ Before opening the draft PR, the branch was rebased onto staging `d444d58541`. T
 
 The follow-up review rebased onto staging `a0f47b1e87`, preserving its 14 updated Search pages, two replaced screenshots and one added screenshot. The compatibility inventory was reconstructed independently from that commit's Git archive using the original Fumadocs AST transform. `baseline.json` records the compatibility commit and retains the original inventory's commit and SHA-256 digest; the initial evidence remains available in commit `3bd99767ac`. The heading conflict preserves the new shared sticky-offset token, which already implements staging's anchor-positioning fix.
 
+The readiness audit merged staging `5facc5f2656fdc2c09b5a5e3b39476ad60f6cb49` at `179c340e8069b59d770cfc763dd42602d8c66c76`. Its compatibility inventory was rebuilt from that exact Git archive with the version-matched Fumadocs heading transform, independently cross-checked against a second Git-blob/MDX inventory. The previous compatibility inventory remains in commit `1aab14faf20b09503d986fbbf8bc2b15c2065e2b` with SHA-256 `b1d6922bf223f0dd79cc4fc097aa1d0a29e20a25def1465b962379e1150c61dc`; the original evidence fields remain unchanged.
+
+The merge preserves seven upstream prose/reference updates, two updated OpenAPI specifications, two replaced Search screenshots, and the replacement of `connect-account-menu.png` with `connect-account-details.png`. These changes document the unified Integrations/account screens, reviewed-action receipts, removal of the extra Search tools gate, the 10,000-iteration condition-loop ceiling, rejected unnamed variable assignments, synchronous workflow wait timeouts and background continuation, and MCP tool output schemas. Refreshing the source inventory keeps these intentional product changes from appearing as rewrite regressions; no preservation assertion was relaxed.
+
+Independent source comparison found no lost authored routes, heading anchors, code fences or operation identities against either main `a52fc509b5b8a18df519ff0781f00c8d45711164` or the new staging commit. Current staging and the merge both contain 495 authored pages, 22,012 headings, 1,236 code fences and 308 operations across seven specifications and 201 paths. All 187 staging assets and all seven specification files match the merge byte-for-byte. The remaining MDX differences from staging are component imports, shared screenshot sizing props, and the Introduction's shared Cards wrapper. The historical reports below were not regenerated by this inventory refresh and do not establish HTTP or browser verification of the new merge.
+
 ## Preserved corpus
 
 - 495 authored MDX documents and 308 generated operations across seven specifications and 201 paths.
@@ -15,7 +21,7 @@ The follow-up review rebased onto staging `a0f47b1e87`, preserving its 14 update
 - The 92 changed MDX files contain only 124 component-import replacements. AST comparison found no other content changes. Workflow graph, configuration, and output fixtures are retained.
 - Existing agent manifest `.mdx` paths and deliberate exclusions remain unchanged. Product-owned OpenAPI, CLI, integration, streaming, and manifest generators remain authoritative.
 
-The full machine-readable accounting is in [report.json](report.json), with the original inventory in [baseline.json](baseline.json).
+Historical machine-readable verification remains in [report.json](report.json). [baseline.json](baseline.json) is the current independently reconstructed compatibility inventory; its provenance points to the original evidence in Git.
 
 ## Results