Skip to content

Latest commit

 

History

History
70 lines (49 loc) · 3.32 KB

CLAUDE.md

File metadata and controls

70 lines (49 loc) · 3.32 KB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Development Commands

Since this is a ComfyUI node collection without a traditional build system, most development is done directly with ComfyUI:

  • Testing nodes: Load ComfyUI with the nodes and test in the UI
  • Dependencies: Install via pip as needed (see StyleTransfer requirements.txt for optional deps)
  • Main dependencies: pip install tensorflow==2.11.0 tensorflow-hub==0.12.0 keras==2.11.0 (for StyleTransfer features)

Repository Architecture

This is a collection of 13 independent ComfyUI custom nodes, each implementing different image processing and utility functions. Each node is self-contained in its own directory following the pattern:

ComfyUI_{NodeName}/
├── __init__.py                  # Node registration (NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS)
└── {implementation}.py          # Main node class with INPUT_TYPES, RETURN_TYPES, FUNCTION

Key Architectural Patterns

Node Registration: All nodes use the standard ComfyUI registration pattern via NODE_CLASS_MAPPINGS and NODE_DISPLAY_NAME_MAPPINGS in __init__.py.

Class Structure: Each node class implements:

  • INPUT_TYPES() classmethod defining required/optional inputs
  • RETURN_TYPES tuple defining output types
  • FUNCTION string naming the main execution method
  • CATEGORY for UI organization

Tensor Conventions: All nodes follow ComfyUI's tensor format [Batch, Height, Width, Channels] with float32 values in 0-1 range.

Node Categories

Image Processing: StyleTransfer (neural style transfer), FaceDetectionNode (OpenCV face detection), channel manipulation nodes Utilities: MemoryOptimizer (GPU/CPU management), RandomSeedGenerator (cross-library seed sync), ImageReloader (dynamic loading) Integration: HoudiniBridge (external 3D software integration) Specialized: NaiveBayesNode (statistical computations)

Version Compatibility

The FaceDetectionNode demonstrates advanced ComfyUI version compatibility - it detects ComfyUI v3 API availability and provides dual implementations, using modern schema definitions when available while falling back to legacy patterns for v1/v2.

External Dependencies

Most nodes use core dependencies (torch, numpy, PIL). Optional dependencies include:

  • opencv-cv2 (face detection)
  • tensorflow + tensorflow-hub (style transfer)
  • scikit-image (enhanced color processing)
  • psutil, watchdog, requests (utility functions)

Node Development Guidelines

When creating new nodes:

  1. Follow the established directory structure and naming conventions
  2. Implement proper error handling and fallback strategies
  3. Ensure tensor format compatibility (BHWC, float32, 0-1 range)
  4. Add proper logging with configurable levels via environment variables
  5. Consider memory management for GPU/CPU operations
  6. Implement IS_CHANGED() method for caching behavior

Integration Patterns

The collection demonstrates several integration approaches:

  • Pipeline Integration: Standard ComfyUI workflow nodes
  • External Tool Integration: Houdini bridge with HTTP API communication
  • Web UI Extensions: Custom JavaScript for enhanced interactions (see reload_image_node)
  • Cross-Platform: Graceful handling of CUDA availability and library dependencies