chore: remove generated deployment and audit evidence

This commit is contained in:
ChatGPT MCP
2026-08-31 05:49:16 +00:00
parent 51c3c90fa1
commit 7d935c5a7d
17 changed files with 0 additions and 2928 deletions
@@ -1,89 +0,0 @@
# MobilityOps premium UI evidence summary
Date: 2026-08-02
Branch: `design/mobilityops-premium-ui`
Baseline revision: `dfabb41582e302f45a3de826f85f531bf23dfc8b`
Final design implementation commit: `1f292e14bb6a2e8ded5dc675b1b3360307d8a9ae`
Review URL: `http://192.168.10.150:1236`
## Outcome
The working PoC was transformed into the Control Rail operational interface without
changing backend contracts or adding scope. All existing journeys remain functional;
return registration gained an evidence-based review boundary before commit.
## Evidence index
- Baseline audit: `docs/design/current-ux-audit.md`
- Three directions and decision: `docs/design/design-directions.md`
- Design tokens and component rules: `docs/design/design-system.md`
- Stitch resource IDs: `docs/design/stitch-manifest.md`
- Implemented visual validation: `docs/design/implementation-validation.md`
- Baseline captures: `artifacts/design-validation/current/`
- Stitch captures: `artifacts/design-validation/stitch/`
- Final responsive captures: `artifacts/design-validation/implementation/`
## Major implementation changes
- Responsive Control Rail shell with compact top bar, desktop rail, off-canvas menu and
labelled mobile bottom navigation.
- Live readiness band, filterable Attention queue, movement timeline, honest integration
pulse and persisted activity on the operations dashboard.
- Searchable fleet and booking registries; booking client pagination limits the DOM to 25
operational rows; responsive tables retain field labels.
- Capture → review → result return workflow with calculated consequence preview and no
write request before confirmation.
- Match/conflict duplicate comparison, evidence-first knowledge, system-health cards and
expandable audit metadata.
- Inline SVG product mark, Feather-like line icon set, CSS control-centre illustration,
timeline/status motion and reduced-motion fallback; no image or motion dependency.
## Validation
| Gate | Result |
|---|---|
| Backend tests | 66 passed |
| Backend lint | ruff passed |
| Backend types | mypy: 0 issues in 44 files |
| Frontend types/build | passed; 59 modules; 240.24 kB JS and 36.63 kB CSS before gzip |
| Browser journeys | 19 passed locally |
| Horizontal overflow | none at 390/768/1280/1440 px |
| Accessibility | named landmarks, skip link, visible focus, text-plus-shape status, labelled mobile rows, reduced-motion support |
| Deployed browser smoke | passed on all 10 authenticated routes plus login at desktop and mobile sizes |
| Console/network | 0 browser warnings/errors; 7 authenticated API paths returned HTTP 200 |
| Deployed global search | Ctrl+K plus `MO-024` navigation passed against the review URL |
All displayed operational counts remain derived from the existing persisted API data.
Synthetic-data labelling is persistent on login and authenticated surfaces.
Deployed evidence is stored in `artifacts/design-validation/implementation/deployed/`.
The review stack reports healthy PostgreSQL/API state, HTTP 200 from the web application,
and healthy state from the server's existing n8n at port 5678. A synthetic return reached
`succeeded` on attempt 1 through that shared n8n and its MobilityOps callback; the demo was
then reset to its deterministic state.
## Performance observations
No runtime font, image or animation dependency was added. The application uses inline SVG
and CSS visuals, and the production bundle remains appropriate for this internal PoC.
## Known limitations
- Global search resolves Control Rail sections and `MO-*`, `BK-*`, `DQ-*` public
references. It intentionally does not offer customer lookup because the locked PoC has
no customer detail route or cross-entity search API.
- The repository retains a bundled n8n service for standalone local clean-checkout demos.
The Unraid overlay keeps it behind the opt-in `bundled-n8n` profile; the live review
deployment uses the server's existing shared n8n instead.
- The MCP Hub state is correctly shown as not configured in the current PoC rather than
simulated as healthy.
- Live RAGcore and MCP Hub round trips remain subject to the existing environment limits
documented in `PROJECT_STATE.md`; their degradation behavior is unchanged.
## Rollback
The accepted baseline remains reachable at commit
`dfabb41582e302f45a3de826f85f531bf23dfc8b`. To roll back the review deployment without
rewriting git history, archive that revision, extract it over the application source on
Unraid while preserving `.env` and Docker volumes, and run
`docker compose -p mobilityops up -d --build`. Verify `/health` and port 1236 afterwards.