Caching, TTLs, and stale-if-error grace

Successful lookups are cached with a multi-day TTL so repeat queries for the same NPI stay fast and reduce load on the upstream registry.

When NPPES is temporarily unavailable, a short grace window may allow serving recently cached data so your workflows degrade gently instead of failing everywhere at once.

Response metadata and headers tell you whether a payload was fresh or served from cache, which helps debugging and customer support.