MyPersona

소개

MyPersona is a Python library and MCP server that augments AI agents with emotional intelligence, enabling them to recall not just what users said, but how they felt. It solves critical gaps in current AI memory systems by providing emotional context, weighting memories based on intensity, and offering a model for behavioral prediction. Its dual-engine architecture, based on Self-Discrepancy Theory, tracks both stated values (Should-Self) and revealed behavior (Want-Self) to predict actions, while naturally fading mundane memories and flagging intense ones for human governance.

주요 기능

• Emotional memory with natural decay (FadeMem) where intense moments persist and mundane ones dissolve
• Subjective Logic for nuanced belief representation distinguishing belief from lack of data
• 0 GitHub stars
• Dual-engine mood tracking and behavioral prediction based on Self-Discrepancy Theory
• Agent introspection for self-modeling, reporting confidence, guesses, and blind spots
• Human-in-the-loop governance for emotionally intense memories requiring approval before storage

사용 사례

• Building AI agents that require emotional context and memory across multiple user sessions
• Detecting discrepancies between a user's stated intentions and their actual behavioral signals
• Creating memory systems that naturally prioritize and retain significant user interactions over time