A real-time GPU cluster monitoring and optimization platform with integrated AI assistant.

Modern AI data centers are the backbone of innovation — but also major energy consumers:

• GPUs burn massive amounts of power, even when idle
• Cooling systems often overcompensate, wasting additional electricity and water
• Inefficient workload scheduling leads to 40%+ energy waste
• By 2030, data centers could consume as much power as entire countries

ThermaMind introduces an AI-driven orchestration layer that continuously monitors and optimizes compute and cooling efficiency in real-time.

• 32 GPU Nodes organized into 4 physical clusters (A, B, C, D)
• Live WebSocket connection updating every 2 seconds
• Heatmap visualization showing GPU load, temperature, and cooling efficiency
• Time-series charts tracking energy usage, cooling efficiency, and workload distribution
• Generates synthetic data for GPU load, temperature, and cooling metrics across all clusters

• AI-optimized cooling: 70% of nodes use predictive cooling (GPU load + 5%)
• Legacy systems: 30% of nodes still catching up from over-cooling issues
• Dynamic status indicators: Active, Idle, or Optimizing based on real metrics
• Predictive node management: Automatically determines when to power down idle nodes and adjust cooling dynamically

• Real-time metrics visualization: Energy savings, CO₂ offset, power usage
• PUE (Power Usage Effectiveness) tracking for efficiency monitoring
• Live performance indicators showing the environmental impact of optimization decisions
• Actionable insights for reducing data center carbon footprint

• Continuous telemetry streams stored in JSONL format (data_logs/telemetry_log.jsonl)
• Ready for future analysis or ML model fine-tuning
• Compact data summaries: Aggregates raw telemetry into hourly insights
• Context-efficient logs optimized for LLM prompts and downstream processing

• Real-time analysis: Ask the AI to analyze current cluster status
• Question answering: Get insights about power consumption, efficiency issues, and workload recommendations
• Text-to-Speech: AI responses are spoken using ElevenLabs voice synthesis
• Context-aware: The AI has access to live telemetry data, not canned responses

• Node.js 18+
• npm or yarn
• Gemini API Key (for AI assistant)
• ElevenLabs API Key (for voice synthesis)

1. Clone the repository:

git clone <repo-url>
cd codeRedAstra

2. Install backend dependencies:

cd backend
npm install

3. Install frontend dependencies:

cd ../frontend
npm install

4. Configure environment variables:

# Copy the example file
cd ../backend
cp .env.example .env

# Edit .env and add your API keys:
# GEMINI_API_KEY=your_key_here
# ELEVENLABS_API_KEY=your_key_here

1. Start the backend server:

cd backend
node src/server.js

The backend will run on http://localhost:8080

2. In a new terminal, start the frontend:

cd frontend
npm run dev

The frontend will run on http://localhost:5173

3. Open your browser to http://localhost:5173

1. Click the 🤖 floating button in the bottom-right corner
2. Click "📊 Analyze Current Status"
3. Wait for the AI to analyze real-time telemetry
4. Read or listen to the AI's insights

Example questions you can ask:

• "Which cluster should I run my next job on?"
• "Why is my power consumption high?"
• "Are there any efficiency issues?"
• "What's happening with Cluster C?"
• "Should I be worried about the cooling on Cluster A?"

• Gemini API: ~$0.001-0.005 per request (flash model)
• ElevenLabs TTS: $0.01 per 1000 characters ($0.005 per response)
• Total: ~$0.006-0.015 per AI interaction with voice

Note: You can disable TTS by removing withAudio: true from the WebSocket messages in AIAssistant.tsx to reduce costs.

• Express server: REST API and static file serving
• WebSocket server: Real-time telemetry broadcast (2-second intervals)
• Simulator: Generates realistic GPU workload and cooling data
• AI Assistant: Gemini integration for telemetry analysis

• React + TypeScript: Modern component-based UI
• Chart.js: Time-series visualization
• TailwindCSS: Responsive styling
• WebSocket hook: Real-time data connection

Simulator → WebSocket Server → Frontend Components
              ↓
         AI Assistant (on demand)
              ↓
         Gemini API → ElevenLabs TTS

backend/
  src/
    server.js          # Express + WebSocket server
    simulator.js       # GPU telemetry generator
    webSocket.js       # WebSocket handlers
    aiAssistant.js     # Gemini + ElevenLabs integration
    
frontend/
  src/
    App.tsx            # Main dashboard
    hooks/
      useTelemetry.ts  # WebSocket connection hook
    components/
      AIAssistant.tsx  # AI chat interface

MIT