Anomix

Anomix is a full-stack behavioral fraud detection platform scaffold.

Stack

• FastAPI backend
• React + Vite frontend
• PostgreSQL database
• synthetic multi-user transaction generator
• XGBoost-ready fraud scoring flow

Docs

• setup: docs/setup.md
• architecture: architecture.md

Backend Database Workflow

Run migrations before starting the API:

cd backend
alembic upgrade head

Training Workflow

From backend/:

python -m ml.generate_dataset --rows 15000
python -m ml.train_model --dataset ml\data\synthetic_transactions.csv

Restart the backend after training so it reloads the saved model artifact.

Recent training metadata now includes:

• train/test row counts
• train/test user counts
• split strategy
• selected decision threshold