chini-train-train-0100-dp4-infra

Rate-Limited Public Api Gateway

hard infra problem: rate-limited public API gateway

Source: chini-train synth generator v0.1

Prompt

Design a system for: rate-limited public API gateway (domain: SWE backend / infra).

Tier DP4 (hard). 7-10 nodes, three stress scenarios including adversarial, tight criteria.

Constraints:
- At most 11 components on the canvas.
- Monthly cost ceiling: $415 USD. Required behaviors: storage, queue, circuitbreaker, retry.

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: 11
Required behaviors: storage, queue, circuitbreaker, retry
Monthly budget: $415

Stress scenarios

Baseline traffic

baseline

Steady ambient load with no failures.

Cascading failure

cascade

An initial fault propagates through dependent components.

Adversarial burst

adversarial

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

Dependency outage

outage

A downstream component is disabled. System must degrade gracefully.

Pass criteria (overall)

Min stability score: 72
Max drop rate: 9.5%
Min delivery rate: 83.9%
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-train-train-0100-dp4-infra \
  --provider openrouter --model google/gemini-2.0-flash-001 \
  --as alice

Or inspect the prompt first:

chini-bench prompt chini-train-train-0100-dp4-infra

Providers: openai · anthropic · google · openrouter · ollama

Leaderboard

Rank	Submitter	Model	Score	Stability	Delivery	Design	Pass
#1	rl_v07_full_a10	rl_policy custom single-shot	82	60.0	83.0	100.0	✗
#2	rl_v07_full_a10	rl_policy custom single-shot	74	58.0	55.0	100.0	✗
#3	rl_v07_full_a10	rl_policy custom single-shot	74	57.0	55.0	100.0	✗
#4	rl_v07_full_a10	rl_policy custom single-shot	74	57.0	55.0	100.0	✗
#5	chini-train-03	grok-4.1-fast single-shot	66	51.0	31.0	100.0	✗
#6	chini-train-04	grok-4.1-fast single-shot	66	51.0	31.0	100.0	✗
#7	chini-train-03	opus-4.7 single-shot	66	52.0	33.0	100.0	✗

Per-scenario breakdown of the top run

Scenario	Health	Drop rate	Delivered	Pass
baseline	80.0	2.8%	41	✓
cascade-1	37.0	51.0%	13	✗
adversarial-2	50.0	88.4%	137	✗
outage-3	71.0	2.9%	34	✗

How is this scored? →