KB-33AF rev 4

UI Preview Master Home — Review Log

6 min read Revision 4
ui-previewmaster-homereview-loggovernance-index

UI Preview Master Home — Review Log

project: ui-preview
surface: master-home
current_version: v1
preview_url: https://vps.incomexsaigoncorp.vn/ui-preview/_master/v1/
status: draft
approval_status: pending_user_review
last_updated: 2026-05-30

v1 — Initial deploy (2026-05-30)

Deployed the static Master Homepage / Governance Index for the UI preview workspace.

  • Data-driven from manifest.json (19 UI surfaces across 4 groups).
  • Summary cards, grouped tables, filters (group/status/approval), search.
  • Featured backlink to MOW Unified Canvas v1 (preview + spec + review-log).
  • Per-row 9-box checklist: idea, cowork_sketch, spec_in_kb, preview_deployed, gpt_reviewed, user_approved, design_polished, code_ready, shell_implemented.
  • No backend calls; loads local manifest.json only.

Gate 0 (verified before deploy)

  • Host = contabo VPS (vmi3080463). OK.
  • /ui-preview/ route works. OK.
  • MOW Unified Canvas v1 → HTTP 200. OK.
  • _demo/health/v1 → HTTP 200. OK.
  • Target _master/v1 was 404 (free). Now 200. OK.
  • No PG / Directus / Qdrant mutation. OK.

Verification (post-deploy)

  • https://.../ui-preview/_master/v1/ → HTTP 200.
  • https://.../ui-preview/_master/v1/manifest.json → HTTP 200, application/json, 19 surfaces.
  • MOW Unified Canvas backlink present.

manifest replace — 61-item Cowork source doc (2026-05-30)

Replaced groups + surfaces in manifest.json from the authoritative Cowork Google Drive source doc, so the Master Home now covers both the DATA layer and the UI layer (previously UI-only, 19 surfaces / 4 groups).

  • Source of truth: Google Drive doc UI-MASTER-GOVERNANCE-61-ITEMS-SOURCE (drive id 1VtOjX0m44taOv9YySKNZ_yV8nKX7YKyKAKa7qv5Bm5g). Read live; not reconstructed from memory.
  • Counts: total 61 = 36 data-layer (registries / tables) + 25 UI-layer (screens).
  • Groups: 16 (9 data groups Nhóm 1–9 + 7 UI groups Nhóm A–G), in source display order.
  • HTML unchangedindex.html not modified. The existing data-driven JS renders the new 16-group / 61-item / two-layer structure as-is (groups iterated dynamically; per-item fields all present). Only manifest.json was replaced.
  • Top-level metadata preserved: schema_version, workspace, generated_by, preview_base, status_vocab, approval_vocab. Added provenance source{} + layers[].
  • Backup: manifest.json.bak-20260530T065336Z (original 19530 B, sha256 62a11741…290e6). New manifest 61223 B, sha256 0554fbb7…7c52f.

Status derivation (per source doc mapping, with one truthful override)

Source mapping: false → not_started; wip → cowork_sketch; true + preview_url=null → spec_ready; true + preview_url set → preview_deployed.

Result: 41 spec_ready, 2 preview_deployed, 2 cowork_sketch, 16 not_started.

  • preview_deployed = 2 — only mow-treeview (A4) and workflow-design-registry (F1), which point at the real, live MOW Unified Canvas v1 (HTTP 200).
  • Truthful override: the source points 5 items — mow-normal-mode (A1), mow-proposal-mode (A2), mot-task-card (B1), moit-form-renderer (C1), mout-table-renderer (D1) — at the v2 URL. …/unified-canvas/v2/ returns HTTP 200 but its own page + preview.meta.json declare it a placeholder scaffold (status: scaffold, approval_status: awaiting_cowork_html, "awaiting Cowork HTML — not agent-designed"). To avoid a false "deployed preview" claim, these 5 are kept at status spec_ready with checklist.preview_deployed=false, their source preview_url (v2) preserved, plus a preview_note flagging the placeholder. They will flip to preview_deployed when Cowork ships real v2 HTML. No fabricated approvals — approval_status is never user_approved.

Verification (post-replace)

  • https://.../ui-preview/_master/v1/ → HTTP 200.
  • https://.../ui-preview/_master/v1/manifest.json → HTTP 200, JSON valid, 61 surfaces / 16 groups / 36 data + 25 UI; deployed-checklist count = 2.
  • MOW Unified Canvas v1 → HTTP 200 (untouched). No other preview dir modified.
  • No PG / Directus / Qdrant mutation.

Source-doc note

The source's aggregate summary ("24 done / 2 WIP / 35 not-started") is internally inconsistent with its own per-row cowork_sketch column (all 36 data rows = true). Status was derived from the explicit per-row values (more granular/authoritative), not the aggregate. Flagged for GPT/User reconciliation.

Open items

  • GPT + User review of the index and the initial status classifications.
  • Many surfaces are not_started with backend present but no UI spec yet — needs Cowork sketches.
  • Replace the v2 placeholder scaffold with real Cowork v2 HTML → then flip A1/A2/B1/C1/D1 to preview_deployed.
  • Decide long-term: keep versioned _master/vN snapshots vs introduce _master/current/ pointer (see manifest-contract.md). v1 chosen for now (no mutable pointer).

Approvals

  • No approvals recorded. approval_status: pending_user_review. No self-approval.