@relkit/cache

Caches declare typed keys, values, logical profiles, and TTL policy. Runtime clients are available only to functions that declare the cache dependency.

import { defineCache } from "@relkit/app/cache";
import { z } from "@relkit/app/schema";

export default defineCache({
  id: "prices",
  profile: "low-latency",
  key: z.object({ sku: z.string() }),
  value: z.number().int().nonnegative(),
  defaultTtlMs: 60_000,
  maxTtlMs: 300_000,
});

Declared functions receive a Promise client with schema-validated keys and values. TTL options use the descriptor policy; numeric value schemas also expose increment.

The package also exposes Effect operations for callers that compose workflows or replace the provider with a test Layer. The Promise client runs the same Effect implementation. Import createCacheClientEffect from @relkit/cache to create a client in an Effect runtime, then run operations such as cache.get("sku") in that runtime.

Cache operations emit cache.<operation> spans and bounded call, failure, and duration metrics. Bridge spans use relkit.cache.<operation> without cache IDs in names or attributes. A CacheTelemetry Layer can replace that observer in tests. Schema and TTL failures retain their exported names, codes, messages, and TypeError or RangeError classifications. Their _tag fields also support typed Effect recovery.