ComfyUI RunPod Package System

A streamlined system for packaging and deploying ComfyUI workflows to RunPod servers. This tool identifies all dependencies in a workflow (models and custom nodes), creates a self-contained package, and provides easy installation on RunPod servers.

Features

• 📦 Interactive Model Discovery: Guided process for identifying and locating models used in workflows
• 🔍 Smart Dependency Detection: Automatically detects custom nodes and their requirements
• 🚀 Efficient Model Downloads: Robust downloading with Civitai API integration and parallel processing
• 🛠️ One-Command Installation: Simple deployment to RunPod servers
• ✅ Installation Verification: Built-in testing to ensure proper setup

Quick Start

# Create a package from a workflow
python example.py create path/to/your_workflow.json --comfyui-dir /path/to/comfyui

# Analyze a workflow without creating a package
python example.py analyze path/to/your_workflow.json

# Test your ComfyUI installation
python example.py test --comfyui-dir /path/to/comfyui

Components

1. Interactive Package Creator

The scripts/interactive_package_creator.py module provides a guided workflow for creating packages:

• Interactively identifies and classifies models used in workflows
• Guides users through locating model files on their system
• Handles large models with options for direct inclusion or download URLs
• Collects Python package requirements from custom nodes
• Generates a standardized package structure with config.json

2. Workflow Parsers

Two workflow parser implementations are provided:

• scripts/simplified_workflow_parser.py: Extracts model references without path resolution
• scripts/workflow_parser.py: Full-featured parser with enhanced model type detection and path resolution

3. Model Downloader

The scripts/model_downloader.py script provides robust model downloading:

• Reliable downloads with retry mechanism
• Civitai API key integration
• Progress reporting with ETA
• Parallel download capability
• Hash verification for integrity

4. Installation Script

The install_package.sh script simplifies deployment to RunPod:

• One-command installation
• Proper handling of requirements.txt
• Efficient model downloading
• GPU optimization settings

5. Main Interface

The example.py script provides a unified command-line interface to all functionality:

• Package creation
• Workflow analysis
• Installation testing

Detailed Usage

Creating a Package

python example.py create path/to/your_workflow.json --comfyui-dir /path/to/comfyui

Options:

• --comfyui-dir PATH: Path to local ComfyUI installation
• --output NAME: Custom name for the output package
• --output-dir DIR: Directory to save the package
• --civitai-key KEY: Civitai API key for model downloading
• --size-threshold GB: Size threshold for large models (default: 2GB)
• --no-manager: Don't include ComfyUI-Manager in the package
• --verbose: Enable detailed logging

Interactive Model Discovery

When creating a package, you'll be guided through the model discovery process:

1. The system will extract model references from your workflow
2. For each model, you'll be prompted to:
   • Select the model type (checkpoint, LoRA, VAE, etc.)
   • Locate the model on your system
3. You'll be given options for large models:
   • Include directly in the package
   • Provide a download URL (recommended for large files)
   • Skip the model

Installing a Package on RunPod

1. Upload the package ZIP file to a web-accessible location
2. On your RunPod server, run:

bash -c "$(curl -sSL https://raw.githubusercontent.com/yourusername/comfyui/main/install_package.sh)" -- --package URL_TO_YOUR_PACKAGE

Replace URL_TO_YOUR_PACKAGE with the URL to your uploaded package ZIP file.

Options:

• --package URL: URL to the package ZIP file (required)
• --dir PATH: ComfyUI installation directory (default: /workspace/ComfyUI)

Package Structure

The created packages follow a standardized structure:

package-name/
├── config.json                 # Package configuration and metadata
├── civitai_config.json         # Optional Civitai API configuration
├── custom_nodes/               # Custom node packages
│   ├── NodePackage1/           # Individual custom node packages
│   └── NodePackage2/
├── models/                     # Model files organized by type
│   ├── checkpoints/
│   ├── loras/
│   ├── vae/
│   └── ...
├── workflows/                  # Workflow JSON files
│   └── workflow.json
├── download_models.py          # Script for downloading external models
└── README.md                   # Package documentation

Development

Extending the System

To add support for new model types:

1. Edit scripts/interactive_package_creator.py
2. Add new model types to the MODEL_TYPES list
3. Update the model type detection logic in the workflow parser

Requirements

• Python 3.8 or higher
• requests library (auto-installed if missing)
• pyyaml library (for parsing extra_model_paths.yaml)

The installation script requires:

• wget, curl, unzip, jq (auto-installed if missing on Ubuntu-based systems)
• Python 3 and pip

Example Workflow

1. Create a workflow in ComfyUI using custom nodes and models
2. Save the workflow as JSON
3. Run python example.py create my_workflow.json
4. Follow the interactive prompts to identify models
5. Upload the resulting ZIP file to a web hosting service
6. Run the installer on your RunPod server
7. Access ComfyUI with your workflow ready to use

For More Information

See SUMMARY.md for a detailed system overview, architecture diagrams, and additional technical details.