# Opt a GET into the response cache

## Goal

Cache idempotent reads at the edge so hot GETs do not consume concurrency slots
or hit the upstream on every call.

## Why

A cache hit skips admission and the upstream fetch, so repeated GETs
do not burn slots. Hits still count toward billing.

Use it when many callers ask for the same idempotent response within a short
window. Skip it for one-off or highly personalized reads where reuse is rare.

## Steps

1. Confirm **Response cache** is enabled on **Pools → Concurrency** (default on).
2. Send an unconditional `GET` or `HEAD` with a TTL:

```bash
curl -sS "https://your-pool.nthpool.cloud/v1/items?page=1" \
  -H "X-Nthpool-Key: $POOL_KEY" \
  -H "X-Nthpool-Cache-TTL: 60"
```

3. Repeat the same request (same path, query, caller identity, and origin).
   Confirm a cache hit in Logs or Metrics.

## Gotchas

- TTL max is **300** seconds.
- HMAC-signed requests are never cached.
- Upstream `Cache-Control: private|no-store|no-cache`, `Set-Cookie`, `Vary`, or
  `206` responses are not stored.
- Hits still count toward billing request allowance.
- Each edge location warms its cache on its own, so the first request in a region is always a miss.

## See also

- [Concurrency](/docs/pools/concurrency) (pool-level response cache toggle)
- [Pools](/docs/pools)