Road to NODES 2026 – Actionable Knowledge With Context Graphs & Agent Memory

Your agent hits the same problem a week later and re-derives everything from scratch. The conversation is retrievable, but the knowledge in it was never structured.

In this live workshop, we use the Neo4j Agent Memory Service (NAMS) to build a context graph that holds short-term, long-term, and reasoning memory in one graph. We design and validate an ontology, import existing models (from Arrows, RDF/OWL, or LinkML, among others), and record every reasoning step and tool call with provenance. Then we distill that history into reusable skills.

You’ll learn how to

See why recall-only memory gives you retrieval without understanding, and how a context graph closes the gap
Design ontologies that limit entity extraction to your domain, and validate them before ingesting anything
Control validation modes, entity-resolution review, and ontology migrations
Capture agent reasoning with provenance and feed it back into context as skills

Trainer

Will Lyon ( linkedin.com/in/lyonwj (opens in a new tab) )