Recon report — liveness, control panel, worlds¶
Top-line + contradictions with the filed record: the engine is ONLINE. Two clarifications to the filed record, neither a failure: (1)
UCCA-TM-2026-07-03-TECH §3lists only the spine — the account holds 18 Workers + 6 Pages projects total; the non-spine ones are UCCA's wider platform (fine) plus RTOpacks residue (below). (2) The retirement brief §A's "docs-proxy" item needs a guardrail: UCCA's owndocs-proxy+knowledge-proxyare live UCCA infra — do NOT retire; the residue ones arertopacks-docs-proxy/rtopacks-proxy(config-only, not deployed).
Unit A — engine liveness: ONLINE¶
| Check | Read | Result |
|---|---|---|
| Gate health | GET ucca-gate…/_health |
HTTP 200 {"status":"ok","surface":"gate"} |
| Proof job | GET /v1/jobs/0dda7509… |
completed, envelope present, seal present |
| D1 job row | SELECT … FROM jobs WHERE job_id=0dda7509 on ucca-input-path |
completed, envelope_ref=default/envelopes/0dda7509….json |
| R2 envelope | envelope_ref resolved + object pulled earlier this session | exists under /envelopes/ |
| Workers deployed | wrangler deployments |
gate + reasoner deployed 2026-07-03; keys deployed today (worker created 2026-03-17) |
| keys reachable | GET keys.ucca.online/v1/public-key |
HTTP 403 = CF managed challenge on the public zone; the worker is up and the spine reaches it via the service binding (not the public route) |
Verdict: ONLINE. The strongest evidence is functional: the proof job is completed with a sealed
envelope, which required all three Workers + D1 + R2 + Queue to work together. Version-ID drift from the
filed values wasn't cleanly extractable via CLI (deployments list showed timestamps, not IDs) — the
deploy dates match today, no drift indicated. No new test job submitted (would burn GLM neurons;
content quality is a separate lane).
Unit B — the control panel: ucca-ops (ops.ucca.online)¶
- Served by the
ucca-opsWorker (ops-v2, Next.js on OpenNext), deployed 2026-03-31.ops.ucca.onlinereturns HTTP 403 +cf-mitigated: challenge(Access/challenge-gated; not read past the gate, as ordered). The permissions-policy header confirms an OpenNext surface behind it. - The world UI/API lives in the ops-v2 repo:
app/w/catalog/worlds/page.tsx,app/api/worlds/…,lib/worlds/rtopacks/api/*. It is a control-panel surface, NOT part of the live engine spine — it talks to world data via its own DB/API, never through the gate. - Full account surface (flagging non-spine): Workers (18) — spine:
ucca-gate,ucca-reasoner,ucca-keys(all 2026-07-03). UCCA platform:ucca-api,ucca-ir,ucca-mcp,ucca-site,ucca-time,ucca-track,ucca-vcc,ucca-corporate,reg-intel,docs-proxy,knowledge-proxy,drive-proxy,ucca-ops(the panel),ucca-backup(live infra). Residue:ucca-backup-large(SUSPENDED). Pages (6):ucca-knowledge,ucca-docs,ucca-trust,ucca-corporate(UCCA);rtopacks-docs,rtopacks-trust(RTOpacks residue).
Unit C — worlds inventory¶
The control panel populates exactly ONE world. lib/auth/seed.sql seeds account_worlds with a
single row: world_id='rtopacks' (full). resolve-auth.ts defaults worlds: ["rtopacks"]. The
account_worlds table (schema.sql) is a real access model, but only rtopacks has data or API code.
The "two more worlds" are engine-repo concepts, not rendered in the panel.
| World | Where it appears | Code? | Data? | Live-spine contact? |
|---|---|---|---|---|
| rtopacks (VET — primary) | engine worlds/rtopacks/ (config); ops-v2 lib/worlds/rtopacks/api/*; account_worlds seed |
Yes (config + ops API) | Yes (panel access row; corpus now in f95d4537) |
No — deployed reasoner excludes worlds/; ops-v2 reaches it via DB/API, not the engine |
| usa_for_later (≈ "US education" — Tim's 2nd) | engine worlds/usa_for_later/ (marketing.py, ucca_config.py) |
Parked fossil (US contextualisation / Home Health Aide, LearnWorlds pivot) | No | No |
| moodle (the 3rd — but NOT a domain world) | engine worlds/moodle/ |
Yes — but its README: "Moodle delivery connector — pushes UCCA capability objects to Moodle as SCORM 1.2 packages." | No | No — a backplane output connector mis-filed under worlds/ |
Tim's open recall — the third world: it is moodle, but the honest reading is that it is a
delivery connector, not a domain world. Only two real domain-world concepts ever existed
(rtopacks + US education); there is no defence/healthcare/aviation/etc. world in code. (worlds/
also holds output/ — generated artefacts, not a world.)
Drift-checks (both applied): no world-shaped structure touches the live spine's contracts. The
deployed reasoner container ships only engine/generator/reasoner (no worlds/); the gate, the v1
schema, and D1 ucca-input-path have no world concept. The ops-v2 "worlds" are a separate control-panel
surface. Verified by the container Dockerfile + the schema + the gate code — not assumed.
Unit D — RTOpacks residue cross-check vs §A¶
| §A item | Live state (read) | Note |
|---|---|---|
| stats-cache, cricos-sync | Not in the 18 deployed Workers | config-only / not deployed — §A accurate |
| qual-enrichment | Not deployed | released by CORRECTION-02 — §A accurate |
| docs-proxy | Ambiguous — GUARDRAIL: UCCA's docs-proxy+knowledge-proxy are live UCCA infra (do NOT retire); the residue is rtopacks-docs-proxy/rtopacks-proxy (repo dirs exist, not deployed) |
§A should name the RTOpacks proxies explicitly |
| ops-v2 / ucca-ops | Deployed (2026-03-31), CUT ruled | §A accurate |
| trust.rtopacks.com.au / rtopacks-trust.pages.dev | Confirmed — Pages project rtopacks-trust present |
§A/C-5 accurate — coordinate with client |
NEW — not in §A: rtopacks-docs Pages project |
present (rtopacks-docs.pages.dev, ~2 months) | RTOpacks residue Pages project — add to the inventory |
NEW — not in §A: ucca-backup-large Worker |
deployed but SUSPENDED 2026-04-09 (UCCA-CLEANUP-01) — "All RTOpacks DB exports removed … Cron disabled. No databases in scope." | already inert; a deploy-surface cleanup, harmless |
§A staleness verdict: substantially accurate; two additions (rtopacks-docs Pages, ucca-backup-large
suspended Worker) and one guardrail (name the RTOpacks proxies, not UCCA's own). Corroboration for the
forensic: ucca-backup-large's own config records a prior UCCA-CLEANUP-01 on 2026-04-09 that
removed the cross-account RTOpacks DB export pattern — consistent with "the DBs left at the 2026-04-08
migration; no later mystery deletion."
HEADs¶
ucca-engine 7d25214 · ucca-docs (this commit). Account e5a98302 throughout; f95d4537 untouched;
nothing deployed, deleted, or changed.