ComfyUI Local Development Setup

A complete ComfyUI template repository with UV package manager for fast dependency management and NVIDIA GPU support. This repository includes all ComfyUI source files for easy deployment on multiple machines.

Prerequisites

• Python 3.12 (recommended)
• UV package manager (pip install uv)
• Git
• NVIDIA GPU with CUDA support (tested with RTX 4080 Super)
• CUDA toolkit installed on your system

Quick Start

1. Clone this repository:

   git clone https://github.com/neuralnetworth/comfyui.git
   cd comfyui

2. Install UV if not already installed:

   pip install uv

3. Install dependencies with UV:

   uv sync

4. Configure ComfyUI Manager for UV support:

   cp comfyui_manager_config.ini custom_nodes/ComfyUI-Manager/config.ini

5. Run ComfyUI:

   # On Unix-like systems
   ./run.sh
   
   # On Windows
   run.bat
   
   # Or directly with Python
   python run.py

   ⚠️ First-Time Startup Note: The initial startup takes 30-60 seconds as ComfyUI-Manager performs security checks and installs dependencies. Subsequent startups will be much faster.

Project Structure

comfyui/
├── .venv/               # UV virtual environment
├── models/              # Model checkpoints
├── input/               # Input images
├── output/              # Generated images
├── custom_nodes/        # Custom ComfyUI nodes
├── .claude/             # Claude AI configuration
├── pyproject.toml       # UV/Python dependencies
├── uv.lock             # Locked dependencies
├── run.py              # Python launch script
├── run.sh              # Unix launch script
└── run.bat             # Windows launch script

GPU Configuration

This setup is configured for NVIDIA GPUs with CUDA 12.8 support. The configuration automatically:

• Detects and uses CUDA on Linux/Windows
• Falls back to CPU on macOS
• Optimizes for NVIDIA RTX 4080 Super

Model Management

Place your Stable Diffusion models in the models/ directory, organized by type:

• models/checkpoints/ - Main model files (.safetensors, .ckpt)
• models/vae/ - VAE models
• models/loras/ - LoRA models
• models/controlnet/ - ControlNet models

You can also configure external model paths in extra_model_paths.yaml.

Development with Claude

This project includes Claude AI integration. Use claude commands for AI-assisted development:

claude --help

Updating ComfyUI

To update ComfyUI to the latest version:

git pull origin main
uv sync

Troubleshooting

CUDA Not Found

Ensure CUDA toolkit is installed and available in your PATH. Verify with:

nvidia-smi
nvcc --version

Module Import Errors

Re-sync dependencies:

uv sync --reinstall

Out of Memory Errors

ComfyUI automatically manages VRAM. For RTX 4080 Super (16GB VRAM), most models should work without issues.

ComfyUI Documentation

For detailed information about ComfyUI's features, supported models, and advanced usage, see README-COMFYUI-ORIGINAL.md.

License

This project setup is open source. ComfyUI itself is licensed under GPL-3.0.