R3
Provides a sub-5ms intelligent memory layer for AI applications, offering fast local caching, real vector embeddings, and knowledge graph capabilities for Large Language Models.
About
R3 serves as an intelligent memory layer specifically designed to enhance AI applications and Large Language Models (LLMs) such as Gemini, Claude, and GPT. It delivers sub-5ms latency through a local-first architecture, leveraging an embedded Redis server for zero-configuration setup. Beyond conventional caching, R3 integrates advanced AI intelligence, including real vector embeddings for semantic search, automatic entity extraction, relationship mapping, and dynamic knowledge graph construction. This allows AI systems to recall information by meaning and context, augmented by multi-factor relevance scoring, while ensuring data reliability with automatic failover to cloud storage.
Key Features
- Semantic search with multi-factor relevance scoring (semantic, keyword, entity, recency, access frequency).
- Automatic failover to cloud storage (e.g., Mem0.ai) for data reliability.
- Sub-5ms intelligent memory layer with fast local Redis caching.
- Zero-configuration setup, local-first operation, and 100% TypeScript for type safety.
- 1 GitHub stars
- Advanced AI Intelligence: real vector embeddings, entity extraction, and knowledge graph construction.
Use Cases
- Integrate intelligent memory into web applications (e.g., Next.js, Vercel AI SDK) to store and retrieve user preferences and context.
- Build and query personal knowledge graphs from stored memories to discover connections and relationships between entities.
- Enhance AI applications and LLM chatbots (e.g., Gemini, Claude) with persistent, context-aware memory for more natural interactions.
