DevOps Incident Triage Agent

An on-call agent that triages incidents by correlating alerts with logs, traces, and runbooks, then proposes next actions.

LangGraphOpenAI APITool CallingKubernetes

Abstract charcoal and magenta cover illustration for the DevOps Incident Triage Agent project, showing correlated alert nodes linked by magenta runbook flows.

Highlights

Tool adapters for PagerDuty, Datadog, Loki, and Kubernetes APIs.
Root-cause hypothesis ranking with cited evidence per hypothesis.
Runbook-aware action suggestions with dry-run and approval gates.

Outcomes

Reduced mean time to acknowledge on tier-1 incidents by 48%.
Auto-resolved 28% of noisy alerts without paging humans.

Stack

LangGraph
OpenAI API
FastAPI
Kubernetes
PostgreSQL