Tasks

These guides walk through the operations you perform against a running inference-cache install — from the first CacheBackend to tuning, isolation, and monitoring.

• Deploy a CacheBackend — the five-minute quickstart: the minimum CacheBackend, what it wires up, and how to read its readiness.
• Bind an engine — the selector → webhook → injection lifecycle, the reserved args/env you cannot override, and its failure modes.
• Tune lookup and eviction — worked CachePolicy examples for match floors, score floors, timeouts, and eviction.
• Isolate tenants — identity, entry-count quota, and the recommended pattern for true per-tenant byte isolation.
• Monitor the cache plane — kubectl surfaces, metrics, and the doctor pre-flight check.