4 cookbooks for multi-tier AI agent caching. LLM responses, tool results, and session state behind one connection - on vanilla Valkey 7+, no modules required.
Install on vanilla Valkey 7+, create an AgentCache, and use all three tiers: LLM cache, tool cache, and session store.
Exact-match LLM response caching with cost tracking. Per-tool TTL policies. toolEffectiveness() recommendations for self-tuning caches.
Per-field TTL, sliding-window refresh, and atomic state updates for multi-step agents. Store intent, reasoning chains, and extracted entities.
BetterDBSaver persists LangGraph graph state on vanilla Valkey 7+ - no Redis 8, no RedisJSON, no RediSearch. Combine with LLM and tool caching for full-stack agent optimization.
Multi-tier agent cache. No modules required. Framework adapters for LangChain, Vercel AI SDK, and LangGraph.