🖥️ AI Station Monitor

A lightweight, real-time web dashboard built with Flask and Docker to monitor NVIDIA GPU performance, system resources, and running Ollama models.

🚀 Features

• GPU Monitoring: Real-time tracking of Fan Speed, Temperature, Power Draw, and VRAM usage via nvidia-smi.
• Ollama Integration: Structured view of currently loaded AI models, including size and processor (CPU/GPU) type.
• System Vitals: Live updates of CPU Load Average, RAM usage, and Disk health (nvme0n1p2).
• Auto-Refresh: Uses JavaScript fetch to update stats every 5 seconds without page flickering.
• Containerized: Fully Dockerized with GPU passthrough and Docker-socket access.

🛠️ Installation & Setup

1. Prerequisites

Ensure your Ubuntu host has the following installed:

• Docker & Docker Compose
• NVIDIA Container Toolkit (Required for GPU passthrough)

2. Clone the Repository

git clone git@github.com:YOUR_USERNAME/ai-station-monitor.git
cd ai-station-monitor

3. Launch the Dashboard

Use Docker Compose to build and run the app in the background:

docker compose up -d

The dashboard will be available at: http://aistation.local:5000 or http://<your-ip>:5000.

---

📂 Project Structure

• app.py: The Flask backend that parses system commands into JSON.
• templates/monitor.html: The frontend UI with CSS Grid and JavaScript auto-refresh logic.
• docker-compose.yml: Handles GPU reservations and mounts the host Docker socket.
• Dockerfile: Builds the Python environment and installs the Docker CLI.

---

🔧 Configuration & Permissions

Docker Socket Access

To allow the dashboard to see your Ollama container status, the container needs permission to read the host's Docker socket:

sudo chmod 666 /var/run/docker.sock

Firewall Setup

If you cannot reach the dashboard from another device, ensure port 5000 is open:

sudo ufw allow 5000/tcp

---

👨‍💻 Development

This project uses Docker Volumes. Any changes you make to app.py or the HTML template will reflect immediately inside the running container without needing a rebuild.

To view live logs for debugging:

docker compose logs -f

📝 License

MIT License - Feel free to use and modify!