GPU Stats Dashboard

Real-time GPU monitoring for DSAI, Rockfish, and IA1 servers. A FastAPI backend runs the collector scripts via SSH and exposes a REST API; a React/Vite frontend displays the data.

Requirements

• Python 3.9+
• Node.js 18+
• SSH access configured for dsai, rockfish, and ia1 hosts

Setup

Backend

# Install Python dependencies (using uv)
uv sync

# Or with pip
pip install fastapi uvicorn

Frontend

cd frontend
npm install

Running

Production (single command)

./run.sh

Builds the frontend (skipped if frontend/dist/ already exists) and serves everything on http://localhost:5000. Pass --rebuild to force a fresh frontend build.

Development (hot reload)

Open two terminals.

Terminal 1 — backend (port 8000):

uvicorn app:app --reload --port 8000

Terminal 2 — frontend (port 5173):

cd frontend
npm run dev

Open http://localhost:5173 in your browser.

API Endpoints

Method	Path	Description
GET	/stats	All servers combined
GET	/stats/dsai	DSAI stats
GET	/stats/rockfish	Rockfish stats
GET	/stats/ia1	IA1 stats
POST	/stats/refresh	Force-refresh all servers
POST	/stats/{server}/refresh	Force-refresh one server

Stats are cached for 15 minutes. The backend refreshes all servers on startup and every 15 minutes in the background.

Project Structure

gpu-stats-ia1-lab/
  app.py                  # FastAPI backend
  pyproject.toml          # Python dependencies
  danielgpus_dsai.py      # DSAI collector script
  danielgpus_rockfish.py  # Rockfish collector script
  danielgpus_ia1.py       # IA1 collector script
  frontend/               # React/Vite frontend
    src/app/App.tsx
    src/app/components/
    src/app/types/