A web application for visually creating machine learning algorithms via drag-and-drop, with reverse engineering to TensorFlow code.

• Drag-and-drop neural network design using ReactFlow
• Reverse engineer visual models to TensorFlow Python code
• Real-time model training with live progress via WebSocket
• CSV and image dataset upload and preprocessing
• Correlation matrix visualization and target field selection

• Node.js >= 18
• Python 3.12+
• PostgreSQL
• Docker (optional, for containerized setup)

# Backend
cd tensormap-backend
docker build -t tensormap-backend .
docker run -p 4300:4300 --env-file .env tensormap-backend

# Frontend
cd tensormap-frontend
docker build -t tensormap-frontend .
docker run -p 3300:3300 tensormap-frontend

cd tensormap-backend
cp .env.example .env          # Configure DB credentials
uv sync
uv run uvicorn app.main:socket_app --reload --port 4300

cd tensormap-frontend
cp .env.example .env          # Optional: configure API URL
npm install
npm start

To enforce code quality, ensure that you set up GitHub Branch Protection rules for the main branch.

1. Go to Settings > Branches.
2. Add branch protection rule for main.
3. Check "Require status checks to pass before merging".
4. Search and select backend-tests and frontend-tests.

To run the backend tests locally with SQLite (fallback):

cd tensormap-backend && pytest tests/ -v

To run tests with PostgreSQL:

export DATABASE_URL=postgresql://test:test@localhost:5432/tensormap_test
cd tensormap-backend && pytest tests/ -v

To run frontend tests locally:

cd tensormap-frontend && npx vitest

tensormap/
  tensormap-backend/    # Python FastAPI server
  tensormap-frontend/   # React + Vite SPA

See CONTRIBUTING.md for guidelines.

Apache 2.0

Oshan Mudannayake

For questions or queries about this project, please reach out via email.