Ontempo · static evidence bundle

Map, QA and polish evidence

A public mirror of the local audit artifacts: the pixel-map QA run, the QA-harness report, the wave-31 engine-polish measurements and the map screenshots. Built so the evidence can be opened without the local stack.

This is not the application. It is a static evidence site. The tenant app needs its Postgres database and Next.js servers; the concrete options for a real public staging app are written up in docs/deploy/live-hosting.md in the repository.

1. Pixel map QA — PASS

Target: http://demo-kelowna-8rs0k.lvh.me:3012/listings?layout=map. The harness captured the real tile layer, decoded every tile with node:zlib, classified it (real / empty / blocked), panned the map and pixel-diffed the before/after frames.

59 / 59real tiles across all captured views
18 / 18real tiles in the landing view
0empty or blocked tiles
PASSpan changed the map (pixel diff)

2. QA harness report

The audit harness inventories every control on a page, runs axe, and flags dead controls and known defects. Numbers below are from the local run against the demo tenant; the JSON/HTML are the source of truth.

115controls inventoried
0dead controls
0axe violations
2findings (1 high, 1 medium)

3. Wave-31 engine polish — before / after

The polish pass fixed shared-sheet and token-level issues across three live tenants. This bundle carries the measured JSON and the curated screenshots (crops of the fixed components plus full-page before/after at 1440 and 390).

4. Interactive map evidence (wave 31 engine)

Screenshots from the engine probe: the map rendered full-page, zoomed in/out, and at 390 px. The probe script and evidence JSON are included for reproducibility.

5. Public staging host — pixel map QA (non-localhost)

The map harness was run against a public Cloudflare host serving the local renderer through a Worker + quick tunnel: ontempo-staging-bridge.website-010.workers.dev/listings?layout=map. The public host served the owner tenant's real page (/listings 200, "36 properties", 8 pins, same-origin /api/map/tiles/… JPEGs).

12 / 12REAL tiles on the public host, 0 blocked
8 / 8pins visible · attribution + same-origin PASS
FAILpan assertion: 0.00% pixels changed on the bridged dev page
6console errors, all the dev HMR WebSocket 502 through the bridge

The public run proves tile authenticity through a real non-localhost host; the pan assertion fails on the bridge because the dev-server page loses its HMR WebSocket through the Worker (local run of the same harness PASSes pan — see §1). Full honesty in docs/deploy/live-hosting.md §4.

What this site proves — and what it does not

ClaimStatus
Real map tiles load and are detected as real (not grey/blocked) on the local appProven by the pixel harness (evidence above)
Panning the map changes the rendered map locallyProven by post-pan pixel diff
The QA harness can inventory controls and run axe on a tenant pageProven by the harness report above
Real tiles are served and classified as real on a public non-localhost hostProven in §5 (12/12 real, 0 blocked, 8 pins)
Map interaction (pan) works through the temporary public bridgeNo — pan check fails on the bridged dev page (HMR socket refused); see §5
The public artifact site is the applicationNo — it is static files only; no tenant routing, no database, no server actions
A real public app deploy existsNo — needs managed Postgres + a Node host or OpenNext adapter; options and effort in docs/deploy/live-hosting.md §6