< Back to Portfolio

Graph Memory Architecture

A memory system that treats context like a graph instead of a chat log.

Entities, relationships, and weights are modeled in Neo4j to enable contextual retrieval through traversal rather than keyword matches.

Relationship strength evolves with recency and interaction frequency, producing a more coherent long-term agent behavior.

Key Highlights

Neo4j graphWeighted traversalContext persistenceAgent memoryResearch prototype