Grid Anomaly Detection

Production-grade anomaly detection for Italian electricity grid data with a React + Node.js frontend and FastAPI backend.

Stack

• React + Vite
• Node.js tooling
• Python / FastAPI
• LSTM + Isolation Forest
• MLflow
• Evidently AI
• Apache Airflow
• Docker
• GitHub Actions
• Vercel-ready dashboard

Structure

• frontend/ - React UI
• src/anomaly_detection/ - backend API and ML code
• airflow/dags/ - retraining workflows
• scripts/ - download/train/check commands
• data/raw/terna/ - downloaded TERNA source data

Run locally

1. npm install
2. npm run dev
3. In another terminal: python scripts\\train_model.py --max-rows 5000 --epochs 1

Vercel

• Deploy the repository as a Vercel project.
• The React app builds from frontend/.
• Set VITE_API_BASE_URL to your hosted FastAPI backend.