chini-005-chat-fanout

Group Chat Fanout (WhatsApp-style)

Deliver messages to large group chats in order. No drops, no duplicates.

Source: Classic system-design interview corpus (WhatsApp / Slack messaging)

Prompt

Design the message-delivery backbone for a group-chat product.

Functional:
- POST /message sends a message to a group of N users.
- Every recipient must receive every message exactly once, in send order.
- Some users are offline and must receive messages when they reconnect.

Non-functional:
- A burst of group messages (5x normal) must not delay 1:1 messages.
- Loss of the offline-storage component must not lose in-flight messages.
- The system must enforce a per-sender rate limit to prevent abuse.

Return a Chinilla CanvasState. You'll need a queue, storage for offline users, fanout to recipients, and a rate-limit on the inbound side.

Constraints

Max components: 13
Required behaviors: queue, ratelimit, storage
Monthly budget: $900

Stress scenarios

Baseline messages

baseline

Normal mix of 1:1 and group chats.

5x group burst

spike

Group-message volume spikes 5x.

Offline storage outage

outage

The storage component for offline users is down.

Noisy network

cascade

Process-time jitter at 30%, ambient 5% drop. Tests resilience.

Pass criteria (overall)

Min stability score: 72
Max drop rate: 3.0%
Min delivery rate: 93.0%
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-005-chat-fanout \
  --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-005-chat-fanout

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	76	65.0	75.0	100.0	✗	X
#2	alex default	X x-ai/grok-4.20	73	48.0	91.0	100.0	✗	X
#3	alex default	A anthropic/claude-sonnet-4.6	71	69.0	57.0	100.0	✗	X
#4	alex default	O openai/gpt-5.4	45	40.0	20.0	100.0	✗	X

Per-scenario breakdown of the top run

Scenario	Health	Drop rate	Delivered	Pass
baseline	68.0	7.7%	168	✗
group-burst	66.0	8.7%	812	✗
offline-storage-down	74.0	3.3%	0	✗
noisy-network	53.0	17.5%	40	✗

How is this scored? →