AI Agent Squad for Ops: One Complex Request, Multiple Agents in Action

1Complex incidents: 5 tabs, 30 minutes of stitching

2am alert: “Why is checkout slowing down?” You need P99, slow traces, root cause, and an incident report.

The old way: Grafana → Jaeger → topology → hand-written summary. Charts everywhere, root cause nowhere — 20–30 minutes per round.

Pain point: Monitoring gives data, not conclusions. Every tool is a separate hunt.

Open-source AI Native OTel APM. OTel ingest + Doris storage + multi-agent troubleshooting.

In one line: orchestrate an AI squad to turn metrics, traces, and topology into an incident-ready report.

curl -fsSL https://databuff.ai/databuff/ai-apm-install.sh | bash

Open http://localhost:27403 (admin / Databuff@123)

Settings → AI model — add your LLM API key:

Demo data: curl -fsSL https://databuff.ai/databuff/ai-apm-demo-install.sh | bash

Drop the §1 request into the AI chat — all 5 screenshots below are from one conversation (live demo).

①

State the goal — P99, traces, root cause, report in one message.

②

Brain dispatches — dispatchExpertTask to metrics + inspection experts.

③

Smart query — breaks down the 240ms entry span.

④

Health inspection — topology + slow traces + metrics to find the bottleneck.

⑤

Deliver summary — remediation + incident report.

vs §1: 5 tabs, 30 minutes → one message, root-cause report in minutes.

No more five browser tabs. Data stays local; code is open and auditable.

Module	What you get
📊 APM	Topology, P99, slow-trace drill-down
🤖 AI squad	Multi-agent queries, synthesized reports
🔌 OTel	Unified trace / metric / log storage

curl -fsSL https://databuff.ai/databuff/ai-apm-install.sh | bash

Try: “Analyze why checkout slowed down and draft an incident report.”