Eight services alerting, one cause: building AI on-call agents on a graph

Description and use

AI agents on a FalkorDB graph accelerate root‑cause identification by querying service dependencies, ownership, recent changes, and past incidents.

When multiple services alert simultaneously, identifying the single root cause is hard because alerts are scattered across unrelated services.

The system reduces root‑cause identification time, eliminates missed causes, and stores incident knowledge for future reuse.

1. Collect alerting component names.
2. Run root‑cause Cypher query to find components common to alerting services.
3. Rank results by coverage, alerting status, dependency alerting, and recent changes.
4. Present top candidates to the investigator agent.
5. Verify candidate with a coverage check; if a better candidate exists, iterate once.
6. Record incident, postmortem, and lesson in the graph.
7. On next incident, use vector‑search + graph to find similar past incidents.

Eight alerts, and only one of them is the actual problem.

None of those answers sits in one place. They're in the connections.

I built the agents on a graph.

The ordering reads the way an SRE thinks: covers, is_alerting, alerting_deps, recent_changes.