OpsMind is an AI-powered incident investigation platform that helps engineers identify the root cause of production outages in seconds.

Instead of manually searching through logs, metrics, deployment histories, and monitoring dashboards, OpsMind automatically analyzes operational telemetry, reconstructs the sequence of events, identifies probable root causes, and generates actionable remediation plans.

Built for modern SRE, DevOps, and Platform Engineering teams.

When a production outage occurs, engineers are often forced to jump between:

• Logs
• Metrics
• Deployment histories
• Monitoring dashboards
• Alerting systems

Finding the answer to a simple question:

What actually happened?

This process can take hours.

OpsMind reduces that investigation time to seconds by transforming operational chaos into structured incident reports.

Production Incident
        │
        ▼
Telemetry Collection
(Log Events, Metrics, Deployments)
        │
        ▼
Evidence Extraction Engine
(Deterministic Analytics)
        │
        ▼
AI Root Cause Investigation
        │
        ▼
Remediation Planning
        │
        ▼
Timeline Reconstruction
        │
        ▼
Final Incident Report

Automatically identifies likely causes of outages using telemetry evidence and contextual reasoning.

Rather than sending thousands of raw logs to an LLM, OpsMind extracts high-signal evidence first:

• Error frequency analysis
• Exception grouping
• Metric anomaly detection
• Deployment correlation

This reduces hallucinations and improves reliability.

Automatically rebuilds the incident timeline:

17:00 Deployment

17:01 Configuration Failure

17:02 Readiness Probe Failed

17:04 CrashLoopBackOff

17:05 Service Outage

Every root cause includes confidence estimates and supporting evidence.

Generates:

• Recommended actions
• Recovery estimates
• Operational risks
• Expected impact

A purpose-built engineering dashboard designed around information density and rapid incident comprehension.

OpsMind uses a hybrid Deterministic + AI architecture.

RAW TELEMETRY
(Logs, Metrics, Deployments)
          │
          ▼
DETERMINISTIC ANALYTICS
(Log Analyzer)
(Metric Analyzer)
(Deployment Correlator)
          │
          ▼
EVIDENCE PACKAGE
(JSON Facts)
          │
          ▼
AI INVESTIGATION LAYER
(Investigator Agent)
(Remediation Agent)
          │
          ▼
VALIDATION LAYER
(Pydantic Models)
          │
          ▼
FINAL INCIDENT REPORT

The deterministic layer extracts objective facts.

The AI layer focuses exclusively on reasoning over evidence rather than raw telemetry.

• FastAPI
• SQLModel
• SQLite
• Pydantic
• Google Gemini
• Async Background Tasks

• React
• Vite
• TypeScript
• Tailwind CSS
• Lucide Icons
• JetBrains Mono

backend/
│
├── app/
│   ├── api/
│   ├── agents/
│   ├── analyzers/
│   ├── builders/
│   ├── database/
│   ├── models/
│   ├── schemas/
│   ├── services/
│   └── main.py
│
├── requirements.txt
└── .env

frontend/
│
├── src/
│   ├── api/
│   ├── components/
│   ├── types/
│   ├── App.tsx
│   └── main.tsx
│
├── package.json
└── vite.config.ts

OpsMind ships with four built-in incident simulations.

Simulates:

• CPU saturation
• Connection pool exhaustion
• Query lock contention
• Service degradation

Simulates:

• Progressive memory growth
• OOMKilled containers
• Pod restart loops

Simulates:

• Third-party dependency failure
• Latency spikes
• Checkout disruption

Simulates:

• Missing configuration values
• Startup failures
• CrashLoopBackOff states

These scenarios enable reliable demonstrations without requiring external infrastructure.

• Python 3.12+
• Node.js 18+
• Gemini API Key

cd backend

python -m venv .venv

# Windows
.venv\Scripts\activate

# Linux / macOS
source .venv/bin/activate

pip install -r requirements.txt

Create .env

GEMINI_API_KEY=your_api_key_here

Run server:

uvicorn app.main:app --reload

Swagger documentation:

http://127.0.0.1:8000/docs

cd frontend

npm install

npm run dev

Application:

http://localhost:5173

1. Open Swagger UI
2. Seed a scenario

POST /demo/seed/database-overload
POST /demo/seed/memory-leak
POST /demo/seed/payment-outage
POST /demo/seed/bad-deployment

3. Open OpsMind
4. Select incident
5. Start Investigation
6. Review:
   • Timeline
   • Root Cause Analysis
   • Confidence Scores
   • Remediation Plan

All AI responses are validated using Pydantic before reaching the frontend.

If AI provider limits are reached, deterministic fallback responses are generated to ensure demo stability.

Existing reports are safely replaced when incidents are re-analyzed.

• Real-time telemetry ingestion
• Datadog integration
• Grafana integration
• Kubernetes event streams
• Historical incident memory
• Automated remediation workflows
• Vector-search powered incident retrieval

HackHazards 2026

AI Incident Root Cause Analyzer for SRE Teams