sailfork / t3code-github / .repos/ alchemy-effect/ website/ src/ content/ docs/ cloudflare/ compute/ cache.mdx

Workers Cache sits in front of a Worker: when it's enabled, Cloudflare checks the cache before invoking the Worker at all, and a hit is served straight from the edge. What gets cached is controlled by standard response headers: Cache-Control (including stale-while-revalidate), Cache-Tag for tag-based purging, and Vary for content negotiation.

Enable the cache on a Worker

yield* Cloudflare.cache() in the Worker's Construction phase. It turns Workers Cache on at deploy time and returns the runtime client. There is no layer to provide:

// src/worker.ts
import * as Cloudflare from "@alchemy.run/cloudflare";
import * as Effect from "effect/Effect";
import { HttpServerRequest } from "effect/http/HttpServerRequest";
import * as HttpServerResponse from "effect/http/HttpServerResponse";

export default Cloudflare.Worker(
  "CachedWorker",
  { main: import.meta.url },
  Effect.gen(function* () {
    const { purge } = yield* Cloudflare.cache();

    return {
      fetch: Effect.gen(function* () {
        const request = yield* HttpServerRequest;

        if (request.url.startsWith("/invalidate")) {
          yield* purge({ tags: ["products"] }).pipe(Effect.orDie);
          return HttpServerResponse.text("purged");
        }

        return HttpServerResponse.text("hello", {
          headers: {
            "Cache-Control": "public, max-age=300",
            "Cache-Tag": "products",
          },
        });
      }),
    };
  }),
);

Responses that carry a cacheable Cache-Control header are stored at the edge; for the next five minutes, repeat requests to the same URL never reach the handler.

Purge by tag

The purge client drives the runtime ctx.cache API from inside a handler. Purge cached responses by the Cache-Tag values they were served with, by path prefix, or wholesale:

yield* purge({ tags: ["products"] });

Failures surface as a typed CachePurgeError:

const result = yield* purge({ tags: ["products"] }).pipe(
  Effect.catchTag("CachePurgeError", (error) =>
    Effect.succeed({ success: false, errors: [error.message] }),
  ),
);

Share the cache across deploys

By default the cache is scoped to a single Worker version, so every deploy starts cold. Opt into sharing cached responses across versions when your responses don't depend on the code that produced them:

const { purge } = yield* Cloudflare.cache({ crossVersionCache: true });

enabled: false keeps the binding's purge client available while turning the read-through cache off, which is useful for toggling cache behavior per stage.

Enable the cache with the cache prop

Workers whose main module exports a plain async fetch handler enable the cache with the cache prop instead:

// alchemy.run.ts
export const Worker = Cloudflare.Worker("Worker", {
  main: "./src/worker.ts",
  cache: { enabled: true, crossVersionCache: true },
});

The handler shapes cache behavior the same way, through Cache-Control, Cache-Tag, and Vary headers on its responses.

Where next