Campaign session

01KWZ5PVXX414YHF04PVWXP97R

finished 2026-07-07 20:53:27.614032+00:00 → 2026-07-07 20:57:27.554021+00:00 · 0 runs · supervisor: react-agent

“Infra-blocked: PostGIS unreachable (127.0.0.1:5432 connection refused) — 0 runs staged, all 3 criteria unresolved; dispatch-layer schema fix recorded for next session.”

Archive snapshot, as of 1 day ago — the run corpus is rebuilt once a day, so this page is not a live reading. The fleet panel is the live one; it refreshes every 30 s.

Success criteria

Criterion	Resolved
Primary: bootstrap-over-floors 95% CI on the LOFO inflation ratio (held-out MAE / within-floor leave-one-occupancy-out MAE) has a lower bound exceeding 1.5x.	no
Variance split: nested decomposition reports sigma^2_floor and sigma^2_seed; geometry-dominates claim holds iff between-floor SD of per-floor CV(N) exceeds within-floor between-seed SD.	no
Comparator honesty: the LOFO denominator is the held-out within-floor baseline (leave-one-occupancy-out), never in-sample training residual.	no

Synthesis

Session synthesis — c-csi-cross-geometry-scaleout (Phase 1 pilot)

Verdict: infrastructure-blocked. Zero runs produced usable artefacts; all three success criteria are UNRESOLVED. No cross_geometry.parquet, no per-floor CV(N) table, no LOFO breakdown, no sigma^2_floor/sigma^2_seed — the artefact contract is entirely unmet.

What happened

Two distinct issues surfaced, one durable-learning, one blocking:

Dispatch-layer schema (resolved, recorded). exp-csi-static uses StaticParamsSchema, which takes n_agents as a scalar int and forbids a seeds field — both are sweep-layer constructs, not single-run params. The correct dispatch is sim_sweep_launch with n_agents and seed as grid axes and campaign constants (experiment, floor, n_placements=48, placement=random, radio.carrier_freq_hz=2.4e9, body.*) pinned as single-value axes. This grid validated cleanly (15 cells/floor) and dispatched — the fix is confirmed and should be reused directly by the next session.
PostGIS unreachable (BLOCKING). Every exp-csi-static cell stages its scene via gis export-floor. The launched multiroom sweep (resplan-12439-floor-0, 15 cells) failed at the first cell with SceneStageError: connection to server at "127.0.0.1", port 5432 failed: Connection refused. This is the exact prerequisite the brief pre-flagged (probed refused 2026-07-06) — still refused 2026-07-07. The remaining 9 floors were not launched: they would fail identically, so committing them would waste budget with no information gain.

Criterion status

C1 (primary inflation-ratio CI): unresolved. Pre-declared under-powered at n=10 anyway; with zero data it is not even borderline.
C2 (variance split sigma^2_floor vs sigma^2_seed): unresolved — the one criterion Phase 1 could have resolved, blocked purely by infra.
C3 (comparator honesty): unresolved — no LOFO computation ran.

Required next action

Restore PostGIS reachability from the runner environment (the runner resolves the DB at 127.0.0.1:5432, not the :5435 the brief anticipated — reconcile the port mapping), verify with gis export-floor on one staged floor, then re-dispatch this identical 10-floor sweep plan. Budget was barely touched (all failures fast, well under the 12 CPU-h cap), so a clean re-run is fully affordable once the DB is up. Figure requests (csi_cross_geometry, seed_trace) are deferred — they require the absent cross_geometry.parquet.

Attached runs

The attached runs are not in the current atlas snapshot — rebuild via web --build-sim-atlas.