Agent Memory Beyond Vectors: Queryable Context Graphs for Reliable AI Systems

Session track: AI Engineering

Session time: 03:30 am - 04:00 am UTC November 12

Session description:

Auxten Wang will show how AI agent memory can move beyond prompt stuffing and vector-only retrieval into a queryable context graph. You will learn how to model tool calls, retrieved documents, user feedback, files, tasks, and outcomes as connected memory that an agent can inspect and reuse. The session will walk through a practical architecture for combining structured event history, semantic retrieval, and graph-shaped context. Auxten will explain where GraphRAG helps, where raw vectors fall short, and how context graphs make agent behavior easier to debug, evaluate, and govern across sessions. You will leave with implementation patterns for building more reliable AI agents: what to store, how to connect it, how to retrieve the right context, and how to measure whether memory is improving outcomes instead of adding noise.