Agentic AI with Knowledge Graphs on the JVM: A Music Graph in Practice

Session Track: Graph Memory & Agents

Session Time: 08:30 am - 09:00 am April 15

Session description

This session continues from Luanne Misquitta’s NODES 2025 talk “Harmonizing AI and Graphs” and focuses on taking a graph-powered application from conversational AI to agentic behaviour.

Using the stepping-stone playlist music application as a running example, the talk shows how to extend a Neo4j-backed system with the Embabel Agent Framework on the JVM. Rather than relying on a single prompt or static tool chains, Embabel allows an agent to dynamically decide which actions to execute to reach a user goal.

You’ll see:
• How a Neo4j music graph acts as the agent’s knowledge graph
• How goals and actions are defined in Embabel
• How the framework sequences and executes actions based on the current context
• How graph queries, LLM calls, and domain logic are combined in a controlled way

The focus is on code, structure, and behaviour, not theory. The session walks through the actual implementation to show how agent frameworks behave in practice when backed by a rich graph model.

This talk is aimed at JVM developers and architects who are already experimenting with AI and want to understand how agents can be built cleanly, predictably, and without turning their application into prompt-driven spaghetti.