012 GitHub stars
02Two Memory Layers: Short-term memory (STM) for verbatim recent messages and Long-term memory (LTM) for extracted facts, decisions, and lessons, both searched simultaneously.
03Context Engine: An active memory system that injects relevant short-term and long-term memories before each AI response.
04Knowledge Graph: Asynchronously connects related memories (decisions, people, rules) to enhance contextual recall and understanding of relationships.
05Five Memory Stores: Organizes memories into `sensory`, `episodic`, `semantic`, `procedural`, and `prospective` categories with distinct retention and search weights.
06Adaptive Recall: Automatically selects from General, Focused, or Deep search modes based on the context of the user's query.
