/setup-comfyui

You are helping the user set up ComfyUI for AI image generation.

Process

1. Check if ComfyUI is already installed

   • Check in ~/programs/ai-ml/ComfyUI (the user's typical location)
   • Look for existing installation

2. Install prerequisites

   • Python 3.10+ (check: python3 --version)
   • Git (check: git --version)
   • For AMD GPU (ROCm):
     • Ensure ROCm is installed: rocminfo
     • PyTorch with ROCm support needed

3. Clone ComfyUI repository

   • Navigate to: cd ~/programs/ai-ml/
   • Clone: git clone https://github.com/comfyanonymous/ComfyUI.git
   • Enter directory: cd ComfyUI

4. Set up Python environment

   • Create venv: python3 -m venv venv
   • Activate: source venv/bin/activate
   • Upgrade pip: pip install --upgrade pip

5. Install dependencies

   • For AMD GPU (ROCm):

     pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.0
     

   • Install ComfyUI requirements: pip install -r requirements.txt

6. Download initial models

   • Create model directories if needed
   • Suggest downloading a base model (SD 1.5 or SDXL):
     • Models go in: ComfyUI/models/checkpoints/
     • VAE in: ComfyUI/models/vae/
     • LoRAs in: ComfyUI/models/loras/
   • Suggest civitai.com or huggingface.co for models

7. Test ComfyUI

   • Run: python main.py
   • Should start on http://127.0.0.1:8188
   • Check logs for GPU detection

8. Create launch script

   • Offer to create ~/programs/ai-ml/ComfyUI/run_comfyui.sh:

     #!/bin/bash
     cd ~/programs/ai-ml/ComfyUI
     source venv/bin/activate
     python main.py
     

   • Make executable: chmod +x run_comfyui.sh

9. Suggest useful custom nodes

   • ComfyUI Manager (for easy node installation)
   • ControlNet nodes
   • Ultimate SD Upscale
   • Efficiency nodes

Output

Provide a summary showing:

• Installation status
• GPU detection status
• Model directory locations
• How to launch ComfyUI
• Recommended next steps (model downloads, custom nodes)
• Troubleshooting tips for AMD GPU