chini-train-train-0190-dp6-infra

Search Autocomplete Service

adversarial infra problem: search autocomplete service

Source: chini-train synth generator v0.1

Prompt

Design a system for: search autocomplete service (domain: SWE backend / infra).

Tier DP6 (adversarial). 11-15 nodes, all scenarios at max, adversarial-heavy. Upper-bound probe.

Constraints:
- At most 14 components on the canvas.
- Monthly cost ceiling: $278 USD. Required behaviors: storage, queue, circuitbreaker, retry, ratelimit.

Return a Chinilla CanvasState that handles the listed scenarios. Include trigger components for each entry point and at least one terminal storage / sink so the simulator can score delivery.

Constraints

Max components: 14
Required behaviors: storage, queue, circuitbreaker, retry, ratelimit
Monthly budget: $278

Stress scenarios

Baseline traffic

baseline

Steady ambient load with no failures.

Dependency outage

outage

A downstream component is disabled. System must degrade gracefully.

Cascading failure

cascade

An initial fault propagates through dependent components.

Traffic spike

spike

Traffic suddenly multiplies. The hot path must hold.

Adversarial burst

adversarial

Hostile packets injected on top of clean traffic. Defenses must block them without dropping good requests.

Pass criteria (overall)

Min stability score: 83
Max drop rate: 5.7%
Min delivery rate: 90.7%
Max errors: 4

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-train-train-0190-dp6-infra \
  --provider openrouter --model google/gemini-2.0-flash-001 \
  --as alice

Or inspect the prompt first:

chini-bench prompt chini-train-train-0190-dp6-infra

Providers: openai · anthropic · google · openrouter · ollama

Leaderboard

Rank	Submitter	Model	Score	Stability	Delivery	Design	Pass
#1	rl_v07_pilot_a10b	rl_policy custom single-shot	74	58.0	56.0	100.0	✗
#2	rl_v07_full_a10	rl_policy custom single-shot	74	57.0	56.0	100.0	✗
#3	chini-train-03	grok-4.1-fast single-shot	68	49.0	45.0	100.0	✗
#4	chini-train-04	grok-4.1-fast single-shot	68	49.0	45.0	100.0	✗
#5	rl_v07_full_a10	rl_policy custom single-shot	67	51.0	37.0	100.0	✗
#6	rl_v07_full_a10	rl_policy custom single-shot	66	50.0	35.0	100.0	✗
#7	rl_v07_pilot_a10b	rl_policy custom single-shot	64	46.0	30.0	100.0	✗
#8	rl_v07_full_a10	rl_policy custom single-shot	64	47.0	32.0	100.0	✗
#9	rl_v07_pilot_a10b	rl_policy custom single-shot	63	45.0	29.0	100.0	✗
#10	rl_v07_pilot_a10b	rl_policy custom single-shot	58	48.0	33.0	75.0	✗

Per-scenario breakdown of the top run

Scenario	Health	Drop rate	Delivered	Pass
baseline	81.0	1.7%	29	✗
outage-1	71.0	2.8%	0	✗
cascade-2	25.0	75.6%	5	✗
spike-3	61.0	16.9%	1364	✗
adversarial-4	50.0	84.5%	127	✗

How is this scored? →