Examples

Search

Clients search a product catalog through a rate limiter, an autoscaled API and a cache in front of a search index. The four building blocks in one system: the limiter turns away what is over its limit, and what it turns away is gone from every box after it; the API adds pods up to its most; the cache answers most searches, so only the misses reach the index; and the clients retry a search the limiter turned away, which past the limit multiplies what it sees. It is laid out stable: no step or control moves a box, since the pool already draws every pod it may run.

Try: ▶ Play: at 30× the retries multiply at the limiter · click Cold cache in the design, and the index takes over · zoom into the API: the faded pods are the ones it may add.

search.loadmap ↓ is the same model as text (on GitHub). Open it in VS Code with the loadmap extension for a live preview beside the text, or in Neovim or Helix: In your editor.

Keeping it true

loadmap check search.loadmap prints search.loadmap: 2 expectations hold, or fails the build that loses one:

expect "Today keeps 10×" { target: 10 }
expect "Proposed keeps 20×" { scenario: Proposed; target: 20 }