Campaign session

01KTPQBFJNBMR8ZNV1Q18VRG3E

finished 2026-06-09 17:37:15.349263+00:00 → 2026-06-09 17:40:34.748400+00:00 · 2 runs · supervisor: react-agent

“Path bug FIXED: the IP-092 structure->sim bridge now carries fiit-library-floor-0 into a JuPedSim seek-and-occupy run end-to-end (exit 0, no DB touch, 0 agents outside walkable, 7 artefacts/run). Occupancy 81% (13/16) >= 0.8; ledger holds (n_seated 13 <= n_goals 16). Caveat: only 16 of 24 authored seats registered as reachable goals and agents clamped 20->16 (=reachable goals) - a seat-reachability/clearance finding (brief's interpretation 2), not a platform bug; seed axis weak (seeds 0/1 byte-identical results).”

Archive snapshot, as of 18 h 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
Each run consumes the staged floor_geometry.json and returns the scalar floor (n_goals, n_seated, occupancy_rate, mean_time_to_seat_s); runner never touches PostGIS	yes
With >= n_agents seats present, occupancy_rate >= 0.8	yes
n_seated <= n_goals and no two agents occupy the same goal	yes

Synthesis

c-fiit-find-a-seat — session 01KTPQBFJN (re-run, path bug fixed)

Re-run rationale. The prior session 01KTGV7KJD (2026-06-07) aborted on a diagnosed singular/plural path mismatch (/work/input/ vs /work/inputs/). That bug is now reconciled in code (floor_walk.py:68, walk_notebook/schema.py:119) and in the brief's sim_params.floor_bundle. This session re-runs the two planned seeds.

Corpus. 2 runs, seeds 0/1, seek_and_occupy, requested 20 agents / target_subtype=seat, on fiit-library-floor-0. Both gate-passed, exit 0, ~11 s wall, 7 artefacts each (sim://run/01KTPQC6D3AXNPHV6XDE71VGJQ, sim://run/01KTPQC7TPYFMQ94T7FPZJKHRS).

Criterion 1 — input contract (✓). Both runs consumed the host-staged floor_geometry.json and returned the full scalar floor (n_goals, n_seated, occupancy_rate, mean_time_to_seat_s); the container never touched PostGIS; n_outside_walkable=0. The path fix holds — no exit-2.

Criterion 2 — occupancy (✓). occupancy_rate=0.812 (13 of 16 reachable seats), ≥ 0.8.

Criterion 3 — ledger (✓). n_seated=13 ≤ n_goals=16; no double-occupancy. Group ledger: spawned 16, seated 13, turned_away 4, departed 0.

Honest caveats (not platform bugs). (a) Only 16 of the 24 authored seats registered as reachable n_goals — 8 seats fall outside the walls-subtracted walkable area at agent_clearance_cm=20 (the brief's interpretation-2 furniture-reachability case); this also clamped agents 20→16 (=reachable goals). The 3 unseated agents are seat-contention/timeout within duration_units=1200, not unreachability (0 agents outside walkable). (b) Seeds 0 and 1 produced byte-identical metrics — the seed axis perturbs only spawn jitter, so the two runs are replicas, not independent draws.

Verdict (brief scenario 1, with a scenario-2 note). The IP-092 bridge carries a real DB floor into a JuPedSim seek-and-occupy run via a portable artefact, all params sim-side — the plumbing is validated. Follow-on: audit why 8 of 24 seats are unreachable (seat placement vs clearance buffer) before occupancy-driven CSI coupling.

Attached runs

Run	Gate	Purpose	Replay
FPZJKHRS	✓	—	replay
DE71VGJQ	✓	—	replay