Cookbooks, demos, and production patterns for semantic caching, vector search, RAG, agent memory, and more.
From caching LLM responses to powering real-time vector search, Valkey is the high-performance backbone for modern AI applications.
Cache LLM responses by meaning, not just exact match. Cut costs by 60%+ and slash latency from seconds to milliseconds.
Scalable, low-latency session storage for chatbot and agent conversations with TTL-based expiry.
Native vector similarity search with HNSW and FLAT indexing. Sub-millisecond nearest-neighbor queries.
Persist tool calls, reasoning chains, and intermediate results for multi-step AI agents with atomic operations.
Token-aware rate limiting for LLM APIs. Sliding window counters and token bucket patterns built in.
Real-time messaging for AI workloads. Broadcast LLM tokens, fan out agent events, and build durable task queues.
Retrieval-Augmented Generation with vector search, document chunking, and metadata filtering - all in Valkey.
Build the unified memory and context layer for AI agents. Assemble context from 5 sources, manage short/long-term memory, prune and budget tokens.
Real-time feature serving for ML models. Microsecond reads for online inference at scale.
Production-ready integrations for popular AI agent frameworks. Checkpointing, caching, vector search, and memory - all backed by Valkey.
Intelligent memory layer for AI agents with a dedicated valkey provider. Per-user memories, vector search, and HNSW indexing.
Use Valkey as the complete persistence layer for LangGraph agents - checkpointing, semantic caching, and vector search - through the official langgraph-checkpoint-aws package.
Give CrewAI agents persistent, searchable memory backed by Valkey. Custom ValkeyStorage backend with GLIDE client, vector search, and Amazon Bedrock embeddings.
Use the official strands-valkey-session-manager community package to back Strands agents with Valkey - persisting conversation history, session metadata, and agent state across invocations.
Use ValkeyDocumentStore and ValkeyEmbeddingRetriever as first-class Haystack pipeline components for sub-millisecond RAG retrieval.
All code is open source. All patterns are production-tested. Clone a repo and start shipping.