Intelligent Neo4j Observability

Session Track: Graph + AI in Production

Session Time: 11:30 am - 12:00 pm April 15

Session description

Observability for graph databases shouldn’t stop at dashboards—it should reason about what it sees. In this talk, I’ll show how Large Language Models and agentic AI can elevate Neo4j operations from reactive monitoring to proactive optimization. We’ll parse logs to understand query intent and behavior, correlate them with the live schema to surface missing indexes and constraint opportunities, and read metrics to spot hotspots, regressions, and capacity risks.

Beyond performance, we’ll use pattern detection to flag security issues—such as risky privileges and configuration weaknesses—then compile findings into clear, actionable reports with concrete remediation steps. Crucially, the system is grounded in Neo4j expert guidance—aligning with official best practices and field-proven recommendations—so suggestions are not just plausible, but correct for Neo4j.

Finally, I’ll demonstrate a “self-healing” loop (only when explicitly authorized) that can test and apply safe changes—like adding targeted indexes or throttling abusive workloads—under guardrails and rollback plans. You’ll leave with a practical blueprint and reference patterns for building an intelligent observability layer for Neo4j that explains problems, suggests fixes, and, when permitted, fixes itself.