Event Memory Module

Overview

The Event Memory module provides episodic storage for significant events in an agent's experience. While the Memory module handles general long-term storage, Event Memory specifically captures and organizes discrete episodes -- notable interactions, milestones, decisions, or occurrences that form the backbone of an agent's experiential narrative.

Episode Detection

Not every interaction warrants an episodic memory. The Event Memory module uses signals from the Awareness module to determine whether the current interaction qualifies as a significant episode. Criteria include emotional intensity, novelty, importance as assessed by the LLM, explicit user requests to remember something, and relationship milestones detected by the Social Network module.

Episode Structure

Each episode is stored as a structured record containing a summary, the full interaction context, extracted entities, emotional valence, importance score, and temporal metadata. Episodes are linked to related entities from the Social Network graph and tagged with topics for efficient retrieval.

Retrieval and Usage

During Context Building, the Event Memory module retrieves episodes relevant to the current interaction. Unlike the Memory module's broad retrieval, Event Memory focuses on significant moments that shaped the agent's understanding. This gives agents the ability to reference past events naturally, recall important conversations, and demonstrate experiential continuity that feels authentic.