One cache API, six stores.

Get, put, remember and atomic locks over memory, Redis, Redis Cluster, Memcached, DynamoDB or Cloudflare Workers KV.

Read the Cache docs

Lookout cache page: 6 hits, 1 miss, 86% hit rate, with each get, put and increment.

What you get

remember and remember_forever for the pattern nearly every cache use is.
Typed reads and writes through CacheExt: get_json::<Post>, put_json.
Atomic locks — the scheduler’s overlap protection uses them.
A store failure the cache works around is a CacheFailed event, not a silent miss.
Honest limit: Workers KV is eventually consistent and has no atomic add.

In the docs

Cache

src/app/http/controllers/stats_controller.rs

use rainier_framework::cache::CacheExt;
use rainier_framework::facades::Cache;

pub async fn totals() -> Result<Response> {
    let posts = resolve::<PostRepository>()?;

    // Computed once, then read from the store for five minutes.
    let totals = Cache::instance()
        .remember("stats.totals", Some(Duration::from_secs(300)), || async {
            posts.totals().await
        })
        .await?;

    Ok(Response::json(&totals))
}

Up close

Select an image to see it full size.

Terminal output: GET /rates takes 44 ms on the first call and under 1 ms after Cache::remember stores it. — remember() runs the slow call once and serves the cached value after that; a failure is never cached.

Telemetry span output for cache.get with cache.hit=false and cache.put. — Every cache call is a span with the store and whether it hit, with no application code.

All features