ADR-0033: Caching Performance Layer

Status

Accepted

Context

InfraLynx now has enough API-owned derived views, CRUD reads, RBAC summaries, and auth/session lookups that repeated uncached reads create avoidable recomputation and repository access.

Decision

InfraLynx will use a centralized cache abstraction backed by Redis, with API-owned keys and explicit invalidation on successful writes.

Consequences

• Redis details stay behind @infralynx/cache-core
• domain packages remain cache-agnostic
• auth, RBAC, inventory, and derived read responses gain bounded cache reuse
• invalidation complexity stays in API write handlers
• stale-data risk is reduced through short TTLs plus targeted invalidation