Weave - Convex + PyTorch Hackathon Project

A modular drag-and-drop neural network editor built for rapid prototyping and experimentation.

Architecture (Hybrid)

1. Frontend (UI): React + Vite + Tailwind + @xyflow/react
2. Backend (Logic & State): Convex (TypeScript) - Handles saving graphs and generating PyTorch code
3. Runner (Execution): Python + FastAPI + UV - Local service that executes generated code

Quick Start

Frontend

cd frontend
npm install
npx convex dev  # Follow the login prompts
npm run dev

Backend (Python Execution Engine)

cd backend
uv sync          # Install dependencies
uv run python main.py  # Start the server on port 8000

Deployment (Netlify + Render + Convex)

This setup deploys the frontend to Netlify, the Convex backend in Convex Cloud, and the FastAPI runner on Render (free tier).

1) Deploy Convex

cd frontend
npx convex deploy

Copy the deployment URL (looks like https://<name>.convex.cloud).

2) Deploy the FastAPI runner on Render

• Create a new Render service from the repo root using the included render.yaml blueprint.
• Set the environment variable ALLOWED_ORIGINS to your Netlify site URL, for example: https://your-site.netlify.app
• Deploy and copy the service URL (looks like https://<service>.onrender.com).

3) Deploy the frontend on Netlify

• New site from Git
  • Base directory: frontend
  • Build command: npm run build
  • Publish directory: frontend/dist
• Set these environment variables:
  • VITE_CONVEX_URL = your Convex deployment URL
  • VITE_RUNNER_URL = your Render service URL

If you use a custom domain, add it to ALLOWED_ORIGINS on Render (comma-separated list).

Tech Stack

• Frontend: React, TypeScript, Vite, Tailwind CSS, React Flow, Lucide Icons
• Backend State: Convex
• Execution Engine: Python 3.11+, FastAPI, PyTorch
• Package Manager: npm (frontend), uv (backend)

MVP Node Types

• Dataset
• Transform
• Model
• Loss
• Optimizer
• Trainer

Project Structure

weave/
├── frontend/           # React + Vite + Convex frontend
│   ├── src/
│   ├── convex/         # Convex functions (auto-created)
│   └── package.json
├── backend/            # Python execution engine
│   ├── main.py
│   ├── requirements.txt
│   └── pyproject.toml
└── README.md

License

Hackathon Project - MIT