Agentmem

关于

Agentmem empowers AI agents with crucial long-term memory, ensuring continuity across sessions where context would otherwise be lost. It stores an agent's knowledge, experiences, and rules in a lightweight, single SQLite file, eliminating the need for complex databases or cloud services. With its innovative hybrid search (combining full-text keywords with vector semantics) and adaptive ranking, agents can quickly recall relevant information. The tool offers robust features like memory tiers, namespaces for isolation, temporal versioning to track fact evolution, and automatic extraction of entities and conversation insights, all designed for production-ready reliability and minimal overhead.

主要功能

• Support for 5 memory tiers (core, learned, episodic, working, procedural) and namespaces
• Multiple operational modes, from zero dependencies to full functionality (0-12MB install)
• 0 GitHub stars
• Hybrid keyword and semantic search with adaptive ranking
• Automated entity extraction and conversation processing (facts, decisions, TODOs)
• Lightweight, persistent memory stored in a single SQLite file

使用案例

• Managing isolated memory stores for multiple AI agents, users, or projects within a single system
• Equipping AI agents with long-term memory that persists across different sessions
• Storing and retrieving an agent's working state, learned knowledge, and behavioral rules efficiently