Cache Management

Every branch and leaf exposes key(). Use these prefix-matchable keys for cache-wide operations:

await queryClient.invalidateQueries({ queryKey: rpcQuery.users.key() })
await queryClient.invalidateQueries({ queryKey: rpcQuery.users.get.key() })

Use queryKey(input) for one payload-specific query:

const input = { id: 1 }
const key = rpcQuery.users.get.queryKey(input)

const cachedUser = queryClient.getQueryData(key)
await queryClient.invalidateQueries({ queryKey: key, exact: true })

Query keys have this flat shape:

;[
  ...keyPrefix,
  'rpc',
  ...rpcTagSegments,
  'query',
  canonicalPayload, // payload-bearing queries only
]

Mutation keys end in 'mutation' and never include variables. See Semantic Keys for normalization and hashing rules.

Manage HTTP caches

HTTP utilities expose the same branch-prefix and ordinary-query key builders. Supply the endpoint’s decoded request input for a specific entry:

const key = http.users.get.queryKey({ params: { id: 1 } })
const cachedUser = queryClient.getQueryData(key)
await queryClient.invalidateQueries({ queryKey: http.users.key() })

HTTP keys include 'http' and the HttpApi identifier after your keyPrefix. RPC keys include 'rpc', so the two factories keep separate caches even with the same caller prefix. Use the original caller prefix when you deliberately want to invalidate both.

The packed RPC consumer and packed HTTP consumer exercise generated keys through QueryClient.