comfYa v0.3.0

ComfyUI Automated Installer for Windows NVIDIA

• Automated Configuration: Automatic detection of hardware profiles and driver capabilities.
• Optimized Execution: Integration of high-performance backend components and kernels.
• Path Portability: Support for dynamic installation paths and environment overrides.
• Verification Framework: Comprehensive system validation and self-healing capabilities.

Requirements

Component	Minimum	Recommended
OS	Windows 10	Windows 11
GPU	RTX 20xx (CC 7.5)	RTX 40xx+
Driver	525+	570+
RAM	16 GB	32+ GB

Quick Start

# Unified CLI commands
.\comfya.ps1 setup
.\comfya.ps1 run
.\comfya.ps1 doctor
.\comfya.ps1 update
.\comfya.ps1 clean

For immediate launch using auto-detected VRAM, use the batch proxy:

.\run.bat

See Quick Start Guide for detailed instructions.

Custom Installation Path

$env:COMFYUI_HOME = "D:\MyComfyUI"
.\comfya.ps1 setup

Project Structure

comfYa/
├── comfya.ps1           # Unified Orchestrator CLI
├── run.bat              # Launch Proxy
├── validate.py          # Installation validator
├── config.psd1          # Central SOT (PowerShell)
├── config.json          # Configuration bridge (Auto-generated)
├── lib/                 # Domain Modules (Lifecycle, Logging, Nvidia, etc.)
├── tests/               # Pester tests
├── docs/                # Documentation
└── .github/workflows/   # CI/CD

Documentation

• Quick Start — Get running in 5 minutes
• Troubleshooting — Common issues and solutions
• Architecture — Technical deep dive
• Contributing — How to contribute

Configuration

Edit config.psd1:

@{
    Version   = "0.3.0"
    Python = @{ Version = "3.12" }
    Cuda = @{ PreferredVersion = "cu128" }
    Gpu = @{ MinComputeCapability = 7.5 }
    LaunchArgs = @{
        HighVram = @("--fast", "--highvram", "--use-sage-attention")
    }
}

Environment Variables

Variable	Description
COMFYUI_HOME	Installation path override
CUDA_MODULE_LOADING	Set to LAZY for optimized CUDA loading
PYTORCH_CUDA_ALLOC_CONF	Memory allocation strategy
TRITON_CACHE_DIR	Triton kernel cache location
TORCH_COMPILE_BACKEND	Compilation backend (default: inductor)
CUDA_PATH	System CUDA installation path (checked by doctor)
HTTP_PROXY / HTTPS_PROXY	Proxy configuration for downloads

Troubleshooting

Common Issues

Issue	Solution
CUDA not detected	Update NVIDIA driver to 570+
Triton import fails	Reinstall: uv pip install triton-windows --force-reinstall
SageAttention missing	See Troubleshooting Guide

Known Limitations

Limitation	Details
Windows only	Linux/macOS not supported (PowerShell + Windows-specific paths)
NVIDIA GPUs only	AMD/Intel GPUs not supported (requires CUDA)
RTX 20xx minimum	Compute Capability 7.5+ required (Turing or newer)
Python 3.11-3.12	Other versions not tested with Triton/SageAttention wheels

Contributing

1. Fork the repository
2. Create a feature branch
3. Submit a pull request

See CONTRIBUTING.md for guidelines.

Changelog

See CHANGELOG.md for version history.

License

MIT