llm-caching-strategies

Caching for LLM applications — provider prompt/prefix caching, semantic response caching, similarity thresholds, scope gates, invalidation. Use when cutting LLM cost or latency. Distinct from redis-caching-patterns key-value caching.