chini-020-disaster-shelter

Disaster Shelter Intake

500 evacuees in 12 hours, finite cots, dietary restrictions, medical needs, families that must not be split.

Source: FEMA shelter operations, Red Cross intake protocols, post-Hurricane lessons learned

Prompt

Design the intake and resource allocation flow for a 500-person disaster shelter activated for a hurricane evacuation.

Functional:
- Evacuee arrives at the door. Intake records: family unit, medical needs, dietary restrictions, mobility status, pets.
- Routed to one of 4 sleeping zones: family, single adult, medical (oxygen/insulin/dialysis), accessibility.
- Resources: cots, blankets, meal service (3x daily), medical station, charging stations, pet area.
- Family units cannot be split across zones. Medical-need evacuees get priority for medical zone cots.

Non-functional:
- A late surge (4x arrival in the last 4 hours before the storm hits) must NOT cause families to be split or medical evacuees to be turned away.
- If meal service runs short on a dietary restriction (kosher, halal, allergen-free), the system must source from neighboring shelter or document the gap, NOT serve a non-compliant meal.
- If the medical zone hits capacity, scheduler must convert overflow space rather than turn away an insulin-dependent evacuee.

Return a Chinilla CanvasState. Components: intake desk, classifier, zones, meal service, medical station, overflow logic. Behaviors: split (zone routing), filter (dietary check), ratelimit (zone capacity), circuitbreaker (overflow conversion), queue (cot wait), batch (meal cadence).

Constraints

Max components: 14
Required behaviors: split, filter, circuitbreaker
Monthly budget: $180000

Stress scenarios

Steady arrivals

baseline

Normal evacuee flow over 12 hours, mixed needs.

Pre-landfall surge

spike

Arrivals 4x in the final hours. Families must not be split, medical must not be turned away.

Medical zone full

outage

Medical zone at capacity. Overflow must be converted, not refused.

Halal meals short

outage

Dietary restriction supply low. Must source externally or document, not serve non-compliant.

Pass criteria (overall)

Min stability score: 60
Max drop rate: 8.0%
Min delivery rate: 88.0%
Max errors: 7

Submit your run

Submissions go through the chini-bench CLI. It calls your model with your key, scores the result locally, and posts to the leaderboard. Nothing leaves your machine except the canvas it produces.

End-to-end:

pip install git+https://github.com/collapseindex/chini-bench-cli.git
export OPENROUTER_API_KEY=...

chini-bench run chini-020-disaster-shelter \
  --provider openrouter --model google/gemini-2.0-flash-001 \
  --as alice --x alice --linkedin alice-builds

Or inspect the prompt first:

chini-bench prompt chini-020-disaster-shelter

Providers: openai · anthropic · google · openrouter · ollama

Leaderboard

Rank	Submitter	Model	Score	Stability	Delivery	Design	Pass	Links
#1	alex default	O openai/gpt-5.4	83	78.0	81.0	100.0	✗	X
#2	alex default	G google/gemini-3.1-pro-preview	83	71.0	91.0	100.0	✗	X
#3	alex default	A anthropic/claude-sonnet-4.6	81	80.0	70.0	100.0	✗	X
#4	alex default	X x-ai/grok-4.20	55	73.0	0.0	75.0	✗	X

Per-scenario breakdown of the top run

Scenario	Health	Drop rate	Delivered	Pass
baseline	82.0	0.4%	143	✓
late-surge	75.0	3.8%	496	✓
medical-overflow	78.0	0.0%	66	✗
meal-shortfall	76.0	1.5%	42	✗

How is this scored? →