Skip to content

Templates

Chinilla comes with 22 templates split into four groups: Interview for practicing system design interviews, Lessons for learning specific behaviors, Examples for real-world technical workflows, and Process for non-technical flows that still benefit from simulation (reasoning chains, ops processes, game systems).

Classic system design interview problems, fully wired up with realistic behaviors, numbers, and costs. Load one, push traffic through it, find the bottleneck, fix it.

Design a TinyURL-style service. Users submit long URLs and get short links back. Way more reads than writes. Watch the cache absorb traffic and the database stay calm.

A real-time messaging system like WhatsApp or Slack. Messages flow through WebSocket connections, get saved, and fan out to recipients.

A multi-channel notification service (email, SMS, push). Events come in, get prioritized, get routed to the right channel, and get tracked for delivery.

A rate limiter that protects APIs from abuse. Watch what happens when traffic goes over the limit: items get throttled, the backend stays healthy.

A file storage service like S3 or Google Drive. Handles big uploads with chunking, metadata indexing, and CDN delivery.

A distributed event streaming platform like Kafka. Publishers send to topics, partitions handle parallelism, subscribers process at their own pace.

A CDN like CloudFront or Akamai. Static content gets cached at the edge close to users. Watch the cache keep the origin server relaxed.

A social media feed like Twitter or Instagram. The classic fan-out problem: when someone posts, who builds the feed and when?

A polite web crawler. The URL queue feeds rate-limited fetchers, the parser pulls out links, the dedupe step drops already-seen URLs, fresh ones loop back into the queue.

Real-time transaction scoring. A rules engine catches obvious fraud, an ML scorer handles edge cases, the decision routes to block / review / approve. Alerts fan out on every block.

Small, focused templates that each teach one idea. Load one, hit play, watch how items flow.

A straight chain: passthrough, filter, queue, delay. Learn how items move through a sequence and where things back up.

If/else routing with weighted split, fan-out across paths, and merging back together. Items take different paths based on the connection weights you set.

Rate limiting, retry with backoff, and circuit breaker working together. Learn how systems survive traffic floods and flaky dependencies. Lines include real travel time (200ms API round-trips) so you can see how delay compounds with retries.

Queue buffering, batch grouping, and replicate broadcasting. See how batching cuts the load on the next block and how replicate copies to every subscriber.

Combines throttling (capacity + serviceRate), weighted split routing, queueing, batch, retry, and replicate in one mini-system. The graduation exercise.

5 blocks wired to show off the circuit breaker: Steady Traffic (trigger) sends items through a Gateway (passthrough, 5ms) to a Payment Service (circuit breaker, 40% failure rate, trip after 2 failures, 3-step cooldown, 200ms processing). Items that survive go into a Result Queue (capacity 20, rate 3) and land in the Order Ledger (storage). The Gateway → Payment line has 50ms latency. Run with different seed counts (N=5 vs N=20) to watch the breaker trip, enter half-open, and probe before recovering.

Evaluate LLM outputs: send prompts through multiple models, score the responses with a judge, send low-confidence answers for human review, log results to a dashboard. Includes a prompt queue, parallel model calls, confidence filtering, and a human review queue for flagged evaluations.

End-to-end pottery production: online orders, throwing, drying rack, glazing, kiln firing in batches, packing, and shipping with inventory tracking. Includes physical travel times between stages.

Classic edge / application / data architecture with subsystems grouped visually. Watch each tier light up with its own boundary throughput as packets flow through the stack.

Non-technical flows where the simulator still earns its keep. These show that Chinilla isn’t only for “tech systems” — anywhere items flow through stages with capacity, retry, and routing semantics.

A claim-evidence-conclusion pipeline. Each claim flows through evidence gathering, counter-arguments (steel-manning), a logic check, and a confidence gate before reaching the final conclusion. Flawed reasoning gets routed to a revision pile. Watch where weak reasoning collapses and how the confidence gate filters out low-quality claims.

Tier-1 to engineering escalation flow. Tickets get triaged by severity (low / medium / high), routed through L1 / L2 / L3 Engineering, retried on miscategorization, and broadcast to both customer notification and CSAT analytics on close. The L3 tier has a circuit breaker so engineering doesn’t get flooded with duplicate reports during outages. Watch the engineering queue back up when severity distribution shifts toward high.

A live-service game’s currency and progression flow. Player actions (combat / quest / craft / explore) earn rewards. Combat and explore drops route through a rarity roll (common 70% / rare 25% / legendary 5%). Every drop is replicated to inventory and telemetry. Watch where progression stalls under different action mixes, or how the legendary drop rate affects long-term engagement.

  1. From the launcher, click Template.
  2. Filter by category (Interview / Lessons / Examples).
  3. Click a card to load it.
  4. Edit, validate, and run as you like.

Start with Interview if you’re prepping for a system design interview. Each template is a complete architecture with realistic behaviors, numbers, and costs. Lessons teach specific behaviors step by step.

The bottom-left of the Pick-a-template modal has two skip-templates actions next to each other:

  • Start blank canvas: empties the modal and drops you onto a blank workspace.
  • Import: opens the native file picker for a previously-exported .chinilla.json (your own work, a teammate’s share, or a translated copy). The full canvas restores: components, connections, drawings, theme, runtime parameters, and a saved sim result if one was bundled.