概要
Engram is an autonomous memory engine designed for AI agents, providing a sophisticated system for knowledge retention and retrieval. It dynamically manages memories through a three-layer lifecycle (Buffer, Working, Core), promoting important information while naturally decaying less relevant data. Memories are also organized into a self-organizing, browsable topic tree based on semantic similarity, with LLM-generated names for easy navigation. This hybrid approach combines time-based decay with space-based clustering, enhancing an agent's ability to recall relevant context, consolidate knowledge, and adapt over time.
