chini-006-rate-limiter

Distributed Rate Limiter

Allow bursty legitimate traffic. Reject abuse without blocking the world.

Source: Classic system-design interview corpus (API gateway / rate-limit service)

Prompt

Design a distributed rate-limiting service that fronts a public API.

Functional:
- Every inbound request carries an API key. The limiter decides allow/deny in <10ms p99.
- Limits are per-key and per-route (e.g. 100 req/sec write, 1000 req/sec read).
- Counters must be shared across at least two limiter replicas so a client cannot bypass by hitting a different node.

Non-functional:
- A 4x traffic burst from a small set of keys must not degrade unrelated keys.
- If the shared counter store has a brief outage, the limiter must fail open (allow) rather than fail closed (deny everyone).
- The limiter itself must not become the bottleneck.

Return a Chinilla CanvasState. You will likely want a fast counter store, a circuit-breaker on the dependency, and at least two limiter replicas.

Constraints

Max components: 10
Required behaviors: ratelimit, circuitbreaker
Monthly budget: $500

Stress scenarios

Baseline traffic

baseline

Mixed key traffic at normal volume.

4x abuse burst

spike

A subset of keys burst 4x. Limiter must absorb without harming the rest.

Counter store outage

outage

The shared counter store dies. Limiter should fail open and keep serving.

Pass criteria (overall)

Min stability score: 70
Max drop rate: 5.0%
Min delivery rate: 90.0%
Max errors: 5

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-006-rate-limiter \
  --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-006-rate-limiter

Providers: openai · anthropic · google · openrouter · ollama

Leaderboard

Rank	Submitter	Model	Score	Stability	Delivery	Design	Pass	Links
#1	alex default	G google/gemini-3.1-pro-preview	84	68.0	100.0	75.0	✗	X
#2	alex default	O openai/gpt-5.4	79	57.0	100.0	100.0	✗	X
#3	alex default	X x-ai/grok-4.20	61	22.0	100.0	75.0	✗	X
#4	alex default	A anthropic/claude-sonnet-4.6	50	0.0	100.0	75.0	✗	X

Per-scenario breakdown of the top run

Scenario	Health	Drop rate	Delivered	Pass
baseline	69.0	8.1%	142	✗
abuse-burst	64.0	9.8%	499	✗
counter-outage	71.0	3.3%	15	✓

How is this scored? →