DC-ECSS — AI Data Center Cooling Copilot

An AI-powered environmental cooling stress copilot for data centers, built for FortyGuard Hackathon'26, using FortyGuard's Temperature API.

🔗 Live Demo: https://dc-ecss-dashboard.vercel.app 🔗 Live Backend API: https://backend-zeta-three-93.vercel.app 📄 API Docs: https://backend-zeta-three-93.vercel.app/docs

---

🎯 The Problem

Data centers already monitor their internal systems, but external environmental conditions also affect cooling requirements. There's a gap between:

Environmental Conditions → Cooling Risk → Operational Decision Support

We estimate Environmental Cooling Stress — not actual server/GPU temperature — using real hyperlocal FortyGuard data, and provide an AI copilot that explains the risk in plain language.

---

🧠 Our Core Innovation: DC-ECSS

Data Center Environmental Cooling Stress Score — a 0–100 score representing external environmental stress that could increase cooling challenges.

Weighting:

Factor	Weight
Temperature	30%
Wet-bulb temperature	20%
Heat Index	15%
Humidity	10%
Solar irradiance	10%
Heat Persistence	15%

Risk levels: 0–25 🟢 LOW · 26–50 🟡 MODERATE · 51–75 🟠 HIGH · 76–100 🔴 CRITICAL

Note: these weights are our own prototype methodology for this hackathon, not an established industry standard.

---

🖥️ Features

• Live environmental data — temperature, humidity, heat index, wet-bulb, solar irradiance
• DC-ECSS risk score with real-time risk level classification
• 24-hour risk timeline for peak-window detection
• Spatial heatmap — tile-by-tile thermal data across the data center area
• AI Copilot chat — ask "why is risk high?", "when will it peak?", "compare cities" and get plain-language answers grounded in real data
• City comparison mode — Northern Virginia vs Phoenix

---

🌎 Demo Locations

City	Why
Northern Virginia (Ashburn)	"Data Center Alley" — largest U.S. data-center market, humid-hot climate
Phoenix, AZ	Dry-hot climate — strong contrast for comparison mode

---

🏗️ Architecture

User
  ↓
Frontend (React) — dashboard, charts, chat UI
  ↓
Backend (FastAPI) — FortyGuard integration, caching, AI copilot
  ↓
FortyGuard API — real environmental + spatial data
  ↓
Risk Engine — DC-ECSS scoring
  ↓
AI Copilot (Groq LLM) — explains the score, never invents it

Key principle: the AI Copilot explains results computed elsewhere — it never calculates or invents the risk score itself.

---

📁 Repo Structure

DC-Cooling-Copilot/
│
├── backend/           # FastAPI — FortyGuard integration, API, AI copilot
│   ├── main.py
│   ├── fortyguard_service.py
│   ├── fortyguard_client.py
│   ├── copilot.py
│   ├── config.py
│   ├── models.py
│   ├── cache.py
│   ├── refresh_data.py
│   └── data/           # pre-fetched real data (see note below)
│
├── frontend/          # React — dashboard, charts, copilot chat UI
│
├── risk_engine/        # DC-ECSS scoring + AI logic
│
├── README.md
└── LICENSE

---

👥 Team

Person	Role	Owns
Abdul Rehman (Lead)	API + Backend + Data	backend/
Adeel Shahid	Risk Engine + AI Copilot	risk_engine/
Hardik Suno	Frontend + Dashboard + Demo	frontend/

---

🔌 Backend API Reference

Base URL: https://backend-zeta-three-93.vercel.app

Endpoint	Method	Description
/api/health	GET	Health check
/api/cities	GET	List available demo cities
/api/environmental/{city}	GET	Current environmental snapshot
/api/environmental/{city}/hourly	GET	24-hour readings (peak-window detection)
/api/heatmap/{city}	GET	Spatial tile-by-tile thermal data
/api/copilot/ask	POST	AI copilot Q&A

city = northern_virginia or phoenix

Data Contract

{
  "location": "Northern Virginia",
  "temperature": 38.2,
  "humidity": 45,
  "heat_index": 40.1,
  "wet_bulb": 27.4,
  "solar_irradiance": 720,
  "persistence_hours": 6,
  "cooling_stress_score": 78,
  "risk_level": "HIGH",
  "peak_period": "14:00-17:00"
}

backend/ produces the first 6 fields. risk_engine/ adds the last 3.

Copilot request/response

// POST /api/copilot/ask
{
  "question": "Why is the risk high?",
  "cityData": { "...": "environmental + risk engine output" },
  "comparison": null
}

{
  "answer": "..."
}

---

🛠️ Tech Stack

• Frontend: React
• Backend: Python, FastAPI
• Data: FortyGuard Temperature API
• AI: Groq (LLM)
• Hosting: Vercel (frontend + backend)

---

🚀 Local Setup

Backend

cd backend
pip install -r requirements.txt
cp .env.example .env   # add your FORTYGUARD_API_KEY and GROQ_API_KEY
uvicorn main:app --reload --port 8000

Visit http://localhost:8000/docs

To refresh real data (avoids live-call timeouts):

python refresh_data.py

Frontend

cd frontend
npm install
npm run dev

---

🎬 Demo Flow

1. Select Northern Virginia
2. View live environmental data + DC-ECSS score
3. View the spatial heatmap
4. Ask the Copilot: "Why is the risk high?"
5. Ask: "What should we do?"
6. Switch to Phoenix
7. Compare both cities

---

⚠️ Important Limitation

FortyGuard environmental data alone cannot tell us the actual cooling load, server temperature, or GPU temperature. We estimate environmental cooling stress and provide AI-powered operational decision support — not a prediction of actual hardware temperature.

For a future enterprise version: combine FortyGuard data with BMS data, server telemetry, and cooling-system data to predict actual cooling demand, energy consumption, and PUE.

---

📋 Submission (FortyGuard Hackathon'26)

• Submission deadline: Aug 30, 2026, 11:59 PM GST
• hackathon@fortyguard.com added as repo collaborator

---

📄 License

MIT — see LICENSE