OPS Agent

An AI-powered incident management system that detects, analyzes, and helps resolve production incidents automatically. Built as a group project to explore multi-agent systems and real-world DevOps automation.

Live Demo: https://ops-agent-flax.vercel.app/

---

What it does

The system simulates a production environment and runs a multi-agent pipeline that handles incidents from detection to resolution:

1. Detects anomalies in system metrics (CPU, memory, latency) and application logs
2. Analyzes the incident using RAG — searches a knowledge base of past incidents to find root causes
3. Notifies via Slack and creates a Jira ticket automatically
4. Waits for approval — a human reviews the proposed fix on the dashboard or Slack
5. Executes the recovery playbook if approved, or escalates to L3 if denied

---

Architecture

The system is split into a React frontend, a FastAPI backend, and a multi-agent pipeline built with LangGraph.

Frontend (React/Vite)
      ↓
Backend (FastAPI)
      ↓
Simulation Engine → Anomaly Detector → Agent Pipeline
                                            ↓
                         Triage → Diagnostics → RCA → Remediation → Comms
                                                ↑
                                        ChromaDB (RAG)

Agents

Agent	Role
Triage	Classifies incident severity and type
Diagnostics	Pulls relevant metrics and log patterns
RCA	Root cause analysis using RAG + Groq LLM
Remediation	Builds a recovery playbook
Communications	Sends Slack alerts and creates Jira tickets
Orchestrator	Coordinates the full pipeline

---

Tech Stack

• Backend — Python, FastAPI, LangGraph, LangChain
• Frontend — React, Vite
• LLM — Groq ( GPT-OSS 120B)
• Vector DB — ChromaDB
• Integrations — Slack, Jira
• Deployment — Render (backend), Vercel (frontend)

---

Running locally

Backend

pip install -r requirements.txt
cp .env.example .env   # add your API keys
uvicorn api.server:app --reload --port 8000

Frontend

cd frontend
npm install
npm run dev

The frontend runs at http://localhost:5173 and proxies API calls to the backend at port 8000.

---

Environment Variables

Variable	Description
GROQ_API_KEY	Groq API key for LLM inference
SLACK_WEBHOOK_URL	Slack incoming webhook for alerts
JIRA_URL	Your Jira instance URL
JIRA_USERNAME	Jira account email
JIRA_API_TOKEN	Jira API token
JIRA_PROJECT_KEY	Jira project key (default: KAN)

---

Features

• Real-time incident dashboard with live metrics and logs
• One-click incident injection for testing (CPU spike, memory leak, service down, etc.)
• Human-in-the-loop approval flow — approve or deny fixes from the dashboard
• Knowledge base viewer with historical incidents
• Incident history with KPIs (auto-resolved rate, average resolution time)
• Slack interactive buttons for remote approvals