Docs gallery: components-disclosure-toggle-group/with-labels names root type "single", which no package registers

[yinlianghui]

Measured while implementing #4616 (the gallery registration widening). Filed unassigned; not fixed there — that PR registers plugin packages, and this entry's red tile is an authoring defect in the entry, not a missing registration. Duplicate-searched (keyword + the toggle-group / single type names): no open issue covers it.

The defect

examples/schema-catalog/src/schemas/components-disclosure-toggle-group/with-labels.json opens with:

{
  "type": "single",
  "items": [
    { "value": "list", "icon": "list", "label": "List" },
    ...
  ]
}

No package in the workspace registers a single component type, so /docs/guide/schema-catalog paints the registry's red panel for this tile:

Unknown component type: single (OBJUI-001)

Its two siblings in the same category spell the same idea correctly — the selection mode is a property, not the type:

• single-selection.json — "type": "toggle-group", "selectionType": "single"
• multiple-selection.json — "type": "toggle-group", "selectionType": "multiple"

So the value that belongs in selectionType ended up in type.

Measured

Rendering all 423 catalog entries the way SchemaThumbnail renders them, against the full registration set #4616 lands, exactly two entries still show OBJUI-001: this one, and core-schema-renderer/unknown-component-type (which demonstrates the panel on purpose). This is the only unintentional one.

Why it was left out of #4616

That card's ruling scopes it to registration and says entry-shape defects are censused and filed rather than rewritten. This entry is named as an explicit exclusion, with this issue number, in examples/schema-catalog/test/catalog-gallery-render.test.tsx — so the pin stays honest about what it does not cover, and removing the exclusion is the natural last step of fixing this.

Fix shape (not prescriptive)

The one-key correction is "type": "toggle-group" plus "selectionType": "single", matching the siblings. Worth confirming first that with-labels is meant to demo the icon+label variant of a single-select toggle group rather than something else, since its items carry icon where the siblings' do not.

Refs #4616, PR #4615.

[yinlianghui]

CLAIM (combined seat: #4624 + #4626 + #4625) — session session_017Qqyix2QcnpUC9XeYVDzx3, branch claude/issue-4624-catalog-authoring-truth. Three catalog-authoring defects, one surface (examples/schema-catalog/src + the #4616 render pin's exclusion list), one seat.

PM ruling (delegated, veto window open):

1. Docs gallery: components-disclosure-toggle-group/with-labels names root type "single", which no package registers #4624: components-disclosure-toggle-group/with-labels — root type single becomes the spelling its two siblings use (toggle-group + selectionType: 'single'), preserving what the entry demonstrates.
2. Docs gallery: the basic-tooltip and basic-hover-card entries render a completely blank tile — each authors a key its renderer never reads #4626: basic-tooltip authors its trigger under the key the renderer reads (trigger/body per the renderer's declaration — read it first); basic-hover-card's value becomes content. Entry meaning unchanged, keys corrected. The CLASS question the filing raises (nothing rejects an undeclared key — the silent-failure mode for AI-authored metadata) is NOT this seat's to solve: leave it recorded on Docs gallery: the basic-tooltip and basic-hover-card entries render a completely blank tile — each authors a key its renderer never reads #4626 for triage; the card stays open for that half if the class discussion deserves its own instrument, else close-with-fix and re-file the class question — seat reports which fits after reading the filing.
3. Docs gallery: the six components-form-calendar entries author the bare "calendar" keyword, which belongs to plugin-calendar's ObjectCalendar #4625 (product ruling): the components-form-calendar category EXISTS to demo the form-calendar primitive, so the six entries should reach the PRIMITIVE, not ObjectCalendar. Verify-first: can a catalog entry address the namespaced registration (ui:calendar) directly as its type? If yes ⇒ re-author the six entries to it (minimal diff, demos preserved) and they leave the pin's DATASOURCE_REQUIRED exemption. If the namespaced type is NOT addressable from schema ⇒ STOP that piece, report the measured options (e.g. a distinct demo-able type name vs registry change — both are contract decisions), and ship pieces 1–2.
4. The Docs gallery: 35 catalog entries render the red "Unknown component type" panel because their plugin package is never loaded on that page #4616 pin's exclusion list shrinks accordingly (its existence-guard case forces the update): fixed entries move from excluded/exempted to fully asserted. Do not touch the pin's other exclusions (Monaco/maplibre environment abstentions stay; the deliberate OBJUI-001 demo stays).

Standard playbook: worktree, red-first per entry (the pin's diagnostics red on the current entries — capture verbatim — green after), reverse verification (no git stash), repo-root vitest (scripts/ examples/schema-catalog), gate battery, no changeset expected (examples ignored — measure; skip-changeset label per #4615/#4628 precedent — labels via union write), DRAFT PR (Fixes #4624, Fixes #4626; #4625 per piece-3 outcome), CI via MCP only (get_check_runs snapshots can be ~60min stale — run-level conclusions via list_workflow_runs are authoritative), no flip/arm.

Mutual exclusion: no other seat in flight. ⛔ content/docs/releases/**; ⛔ apps/site registration (landed, not this card); ⛔ renderer-side changes (key-rejection class = future instrument).

---

Generated by Claude Code

---

Generated by Claude Code

[yinlianghui]

PM step-7 复核: ACCEPT — both deviations ruled in the seat's favor. (1) The hover-card substitution stands: the ruled value→content swap was measured a render-time no-op that would also author a key TextSchema does not declare — the trigger-element fix is what achieves the ruling's intent (the hover card can actually open). (2) #4626 closes with this fix; its class question lives on as #4631 with strictly stronger specimens (a REQUIRED children the renderer never reads). PR #4630 flipped and armed. #4631/#4632/#4633 stand as filed for the next PM pass.

---

Generated by Claude Code

---

Generated by Claude Code