TypedMem

A long-term memory management system designed for AI agents. Rather than relying purely on a retrieval paradigm, it employs a state-evolution model and introduces explicit parsers to handle supersession relationships between pieces of information, with support for temporal validity and preference-conflict resolution.