verygoodplugins/automem
📦 Open Source Projectverygoodplugins
A graph-vector memory service providing AI assistants with durable, relational long-term memory.
AutoMem addresses the critical limitation of stateless LLMs by providing a robust, persistent memory layer. Unlike standard vector-only RAG implementations, AutoMem leverages a hybrid approach: it uses vector databases (Qdrant) for semantic search and graph databases (FalkorDB) to map complex, relational data between entities. This dual-structure allows AI agents to perform sophisticated reasoning over historical data, such as understanding the connections between different users, projects, or past decisions. The service is built in Python and designed for easy integration into existing AI agent workflows. It supports advanced memory management, allowing developers to define how information is stored, updated, and retrieved, ensuring that the AI's 'knowledge' evolves alongside the user's interactions. By offloading memory management to a dedicated service, developers can focus on agent logic while AutoMem handles the heavy lifting of state persistence and relational context retrieval.
💡Highlights
- ├─Hybrid graph-vector memory storage
- ├─FalkorDB and Qdrant integration
- └─Persistent relational context
🎯For
- ├─AI Engineers
- └─Backend Developers