> ## Documentation Index > Fetch the complete documentation index at: https://docs.usecroma.com/llms.txt > Use this file to discover all available pages before exploring further. # Live lookups > How Croma live lookups answer: straight from the source at request time, with the source's own freshness, and async delivery when they run long. Most endpoints query the source when you call: these are **live lookups**. The answer comes from the source at that moment, in the same JSON shape as every other endpoint. The rest are [dataset endpoints](/datasets), and any endpoint without the **Dataset** pill in its guide is live. A live lookup: * returns what the source says right now, not as of a refresh, * resolves one subject per call (a cédula, a plate, a radicado), * spends **10 credits** per request (see [Rate limits](/rate-limits)). ## Latency Most live lookups answer in a few seconds, at the pace of their source. The slowest run as [async jobs](/async-jobs): wait inline, poll, or get a callback; that page lists which endpoints they are. A repeated identical lookup can answer from a recent cached result (`X-Cache: HIT`). Cached answers spend the same credits as a miss. ## Availability A live lookup is only as available as its source. When the source fails or is unreachable, the endpoint returns a `5xx` with `type: "upstream_error"` (see [Errors](/errors)). Lookups are idempotent, so retrying is always safe. The other answer mode: the whole source, queried in milliseconds.