Building a Graph-Vector Hybrid Memory Layer for AI Agents

I built a memory system for AI agents that distills task failures into single actionable rules — stored in ChromaDB and linked conceptually in a Neo4j graph — instead of dumping raw conversation logs. Includes rule decay, multi-tenancy, and an honest look at what's still unranked in retrieval.