⚡ AI Workstation Monitor

A high-performance, single-file hardware monitoring dashboard designed for Deep Learning Engineers, Data Scientists, and Power Users.

Stop using watch -n 1 nvidia-smi. Get a professional, real-time visual cockpit of your workstation directly in your browser.

🚀 Why this monitor?

Most monitoring tools are either too simple (CLI based) or too heavy (Prometheus/Grafana stacks). This project bridges the gap:

• Zero-Lag: Uses lightweight HTML5 Canvas and Flask. No React/Vue overhead.
• NVIDIA Native: Deep integration with pynvml for precise GPU metrics (Power, Fan, VRAM, PCIe).
• Process Inspector: Instantly see who is hogging your RAM or VRAM.
• Modern Stack: Built with modern Python tooling (uv).
• Robust: Auto-detects hardware. Falls back to "CPU Mode" gracefully if no NVIDIA GPU is found.

✨ Features

• Real-time GPU Telemetry: Load, VRAM, Temperature, Power Draw (Watts), Fan Speed, PCIe Throughput.
• System Metrics: CPU Load (Global & Per-Core), RAM Usage, Swap Memory, SSD Storage.
• Top Processes: Live table of the most resource-intensive processes (CPU/MEM).
• Dark Mode UI: "Cyberpunk/NVIDIA" aesthetic designed for dark environments.

📦 Installation

This project uses uv for ultra-fast dependency management.

Prerequisites

• NVIDIA Drivers installed (for GPU metrics).
• uv installed.

1. Clone the repository

git clone https://github.com/hasso5703/neurodash.git
cd neurodash

2. Sync dependencies

Initialize the environment and install dependencies in milliseconds:

uv sync

🚀 Usage

Development / Quick Start

Launch the monitor directly within the managed environment for testing or personal use:

uv run main.py

Access: http://localhost:9999

Production Deployment (Recommended)

For long-running AI workstations, use Gunicorn to ensure stability and performance without blocking resources.

Option A: Manual Launch Use this command to run a single worker with multiple threads. This minimizes RAM usage while keeping the interface responsive.

uv run gunicorn -w 1 --threads 4 --worker-class gthread -b 0.0.0.0:9999 main:app

Option B: Auto-Start on Boot (Linux/Ubuntu) 🐧 Want the dashboard to start automatically when your workstation turns on? We provide a one-line installer that sets up a systemd service for you.

1. Run the installer:

   chmod +x install_service.sh
   ./install_service.sh

2. That's it! The service will now run in the background and restart automatically if it crashes.

Manage the service:

• View logs: sudo journalctl -u neurodash -f
• Stop: sudo systemctl stop neurodash
• Restart: sudo systemctl restart neurodash

🌐 Remote Access

Since the monitor binds to 0.0.0.0, it is accessible from any device on your network.

🏠 Local Network (WiFi/LAN)

Access the dashboard from your phone or tablet connected to the same WiFi/Ethernet: http://<YOUR_WORKSTATION_IP>:9999

(Example: http://192.168.0.29:9999)

Tip: Run ip addr or ifconfig in your terminal to find your local IP address.

🌍 Anywhere Access (Tailscale)

For secure remote monitoring outside your home (e.g., checking training status from a cafe), we recommend using Tailscale.

1. Install Tailscale on both your workstation and your mobile device.
2. Access the dashboard securely via your workstation's Tailscale IP: http://<TAILSCALE_IP>:9999

This method is secure (VPN) and does not require opening ports on your router.

🔧 Configuration

The application is configured via environment variables or default values in main.py:

Variable	Default	Description
HOST	0.0.0.0	Listen on all interfaces
PORT	9999	Web server port
UPDATE_INTERVAL	1000	Refresh rate in ms

🛠️ Troubleshooting

"No GPU Detected" / CPU Mode: If the dashboard runs but shows "No GPU", ensure you have NVIDIA drivers installed. The tool gracefully handles machines without GPUs (e.g., MacBooks, Laptops) by hiding the GPU panels.

Missing Libraries: If nvidia-ml-py fails to load (e.g. on non-Linux systems), the script will automatically switch to CPU-only mode.

📄 License

MIT