Agentic AI with Knowledge Graphs on the JVM: A Music Graph in Practice
summary
A Neo4j-backed personal record graph plus embeddings and GDS powers two JVM agents built with the Embabel Agent Framework: a chatty discovery agent that recommends or adds albums and an autonomous path-building agent that stitches five listening “parts” using GDS Yen shortest paths, MusicBrainz/Postgres fetches, LLM summaries, and similarity re-projections. Planning is deterministic (GULP), agent selection hits ~95% confidence, loops are limited to two expansion attempts, and guardrails (blocklist, budget) stop hallucinations—overall it works but required those limits to avoid runaway graph changes.
AI-written summary · about
chapters (8)
- 0:00Introduction and motivation
- 5:02Music graph structure explained
- 10:21Discovery agent flow overview
- 14:14Code walkthrough of discovery agent
- 17:03Collection expansion demo
- 19:51Autonomous listening path agent
- 23:01Complex path demo and evaluation
- 27:37Key takeaways and conclusions
related talks
32mHarmonizing AI and Graphs: Building Intelligent Music Apps with Neo4j and AI
20mEnhancing RAG with Multi-Agent Integration
14mEmpowering Voice AI Agents with Real-Time Knowledge from Neo4j
54mGenAI Grows Up: Building Production-Ready Agents on the JVM
08mHarness Graph Power With Agentic AI: An Insights Discovery Use Case
30mAgentic RAG with Knowledge Graphs Explained
28mData-Empowered AI Assistants
29mNext-Gen Travel Agents: Merging AI Agentic and Neo4j for Personalized Experiences