nodes.py

import torch
from comfy.utils import lanczos

class AutoImageResize:
    @classmethod
    def INPUT_TYPES(s):
        return {
            "required": {
                "image": ("IMAGE",),
                "desired_width": ("INT", {"default": 1024, "min": 64, "max": 8192, "step": 8}),
            }
        }

    RETURN_TYPES = ("IMAGE", "INT", "INT")
    RETURN_NAMES = ("image", "width", "height")
    FUNCTION = "execute"
    CATEGORY = "finetuners"

    def execute(self, image, desired_width):
        # Get current dimensions
        _, current_height, current_width, _ = image.shape
        
        # Calculate target width and scale factor
        target_width = current_width
        if current_width < 1024 or current_width > 1344:
            target_width = desired_width
            scale_factor = desired_width / current_width
        else:
            # No resize needed
            return (image, current_width, current_height)
            
        # Calculate new height maintaining aspect ratio
        target_height = int(current_height * scale_factor)
        
        # Convert to NCHW for lanczos
        x = image.permute(0, 3, 1, 2)
        
        # Perform lanczos resize
        x = lanczos(x, target_width, target_height)
        
        # Convert back to NHWC
        x = x.permute(0, 2, 3, 1)
        
        return (x, target_width, target_height)


class GroupLink:
    @classmethod
    def INPUT_TYPES(s):
        return {"required": {}}
    
    RETURN_TYPES = ()
    FUNCTION = "noop"
    CATEGORY = "finetuners"

    def noop(self):
        return {}


class VariablesInjector:    
    @classmethod
    def INPUT_TYPES(s):
        return {
            "required": {
                "Var1": ("STRING", {"forceInput": True}),
                "prompt": ("STRING", {"multiline": True, "height": 4})
            },
            "hidden": {
                "Var2": ("STRING", {"forceInput": True}),
                "Var3": ("STRING", {"forceInput": True}),
                "Var4": ("STRING", {"forceInput": True}),
                "Var5": ("STRING", {"forceInput": True}),
                "Var6": ("STRING", {"forceInput": True}),
                "Var7": ("STRING", {"forceInput": True}),
                "Var8": ("STRING", {"forceInput": True})
            }
        }
    
    RETURN_TYPES = ("STRING",)
    RETURN_NAMES = ("text",)
    FUNCTION = "inject"
    CATEGORY = "finetuners"
    
    def parse_var_input(self, input_str):
        """Parse input string in format 'name | value'"""
        if not input_str:
            return None, None
            
        parts = input_str.split("|", 1)
        if len(parts) == 2:
            return parts[0].strip(), parts[1].strip()
        return None, input_str.strip()
    
    def inject(self, prompt, **kwargs):
        result = prompt
        
        # Process each var input
        for i in range(1, 9):
            var_key = f'Var{i}'
            var_input = kwargs.get(var_key)
            
            if var_input:  # Only process if we have a value
                name, value = self.parse_var_input(var_input)
                if name:  # Only inject if we got a valid name
                    result = result.replace(f"!{name}", str(value))
        
        return (result,)

# Node mappings
NODE_CLASS_MAPPINGS = {
    "VariablesInjector": VariablesInjector,
    "AutoImageResize": AutoImageResize,
    "GroupLink": GroupLink
}

NODE_DISPLAY_NAME_MAPPINGS = {
    "VariablesInjector": "🔄 Variables Injector",
    "AutoImageResize": "📐 Auto Image Resize",
    "GroupLink": "🔗 Group Link"
}

# Informs user that nodes are loaded
print("\033[34mComfyUI Finetuners: \033[92mLoaded\033[0m")