cache-guide

Use cache() to cache entire routes or segment trees with runtime conditions and custom keys, and use "use cache" to cache individual function results keyed by arguments. cache() skips the whole handler on hit; "use cache" skips only the wrapped function body.

Add the "use cache" directive inside the query function, optionally with a profile like "use cache: short". The cache key includes the function identity and serialized non-tainted arguments, so different arguments produce separate entries.

No, loaders are always excluded from route-level cache() and resolve fresh on every request, even on a full segment cache hit. To cache a loader's data, opt in explicitly with loader(Fn, () => [cache({ ttl: ... })]).

They throw because per-request data is not part of the cache key, so reading them inside a cached scope would bake one request's data into a shared entry. Response side effects like ctx.header() are also blocked since they would be lost on a cache hit.

CFCacheStore prefixes every store key with the build version, so a new deploy never reads a previous build's entries. This prevents cross-deploy shape drift but means every deploy starts with a cold data cache.

No, the outer cache window bounds everything inside it. On an outer cache hit no inner code runs, so an inner shorter TTL only takes effect when the enclosing cache recomputes; use a loader for data that must stay fresher.