all new nodes working under EXR tab category

README.md

@@ -2,8 +2,6 @@
 
 Open source comfyui deployment platform, a `vercel` for generative workflow infra. (serverless hosted gpu with vertical intergation with comfyui)
 
-Check out our latest lcoal demo -> https://github.com/comfy-deploy/comfyui-api-comfydeploy
-
 > [!NOTE]  
 > Im looking for creative hacker to join ComfyDeploy's core team! DM me on [twitter](https://x.com/BennyKokMusic)
 

comfy-nodes/external_exr.py

@@ -1,110 +0,0 @@
-import os
-import io
-import cv2 as cv
-import numpy as np
-import torch
-import requests
-from folder_paths import get_annotated_filepath
-
-class ComfyUIDeployExternalEXR:
-    RETURN_TYPES = ("IMAGE", "MASK")
-    RETURN_NAMES = ("image", "mask") 
-    FUNCTION = "load_exr"
-    CATEGORY = "🔗ComfyDeploy"
-    
-    @classmethod
-    def INPUT_TYPES(cls):
-        return {
-            "required": {
-                "input_id": (
-                    "STRING",
-                    {"multiline": False, "default": "input_exr"},
-                ),
-                "exr_file": ("STRING", {"default": ""}),
-                "tonemap": (["linear", "sRGB", "Reinhard"], {"default": "sRGB"}),
-            },
-            "optional": {
-                "default_image": ("IMAGE",),
-                "default_mask": ("MASK",),
-                "display_name": (
-                    "STRING", 
-                    {"multiline": False, "default": ""},
-                ),
-                "description": (
-                    "STRING",
-                    {"multiline": False, "default": ""},
-                ),
-            }
-        }
-
-    @classmethod
-    def VALIDATE_INPUTS(s, exr_file, **kwargs):
-        return True
-
-    def sRGBtoLinear(self, npArray):
-        less = npArray <= 0.0404482362771082
-        npArray[less] = npArray[less] / 12.92
-        npArray[~less] = np.power((npArray[~less] + 0.055) / 1.055, 2.4)
-
-    def linearToSRGB(self, npArray):
-        less = npArray <= 0.0031308
-        npArray[less] = npArray[less] * 12.92
-        npArray[~less] = np.power(npArray[~less], 1/2.4) * 1.055 - 0.055
-
-    def load_exr(self, input_id, exr_file, tonemap="sRGB", 
-                 default_image=None, default_mask=None,
-                 display_name=None, description=None):
-        try:
-            if exr_file and exr_file != "":
-                if exr_file.startswith(('http://', 'https://')):
-                    # Handle URL input
-                    response = requests.get(exr_file)
-                    # Write to temp buffer
-                    buffer = io.BytesIO(response.content)
-                    nparr = np.frombuffer(buffer.getvalue(), np.uint8)
-                    image = cv.imdecode(nparr, cv.IMREAD_UNCHANGED).astype(np.float32)
-                else:
-                    # Handle local file
-                    exr_path = get_annotated_filepath(exr_file)
-                    image = cv.imread(exr_path, cv.IMREAD_UNCHANGED).astype(np.float32)
-
-                if len(image.shape) == 2:
-                    image = np.repeat(image[..., np.newaxis], 3, axis=2)
-                
-                # Extract RGB and flip channels
-                rgb = np.flip(image[:,:,:3], 2).copy()
-                
-                # Apply tonemapping
-                if tonemap == "sRGB":
-                    self.linearToSRGB(rgb)
-                    rgb = np.clip(rgb, 0, 1)
-                elif tonemap == "Reinhard":
-                    rgb = np.clip(rgb, 0, None)
-                    rgb = rgb / (rgb + 1)
-                    self.linearToSRGB(rgb)
-                    rgb = np.clip(rgb, 0, 1)
-                
-                rgb = torch.unsqueeze(torch.from_numpy(rgb), 0)
-                
-                # Handle alpha/mask
-                mask = torch.zeros((1, image.shape[0], image.shape[1]), dtype=torch.float32)
-                if image.shape[2] > 3:
-                    mask[0] = torch.from_numpy(np.clip(image[:,:,3], 0, 1))
-
-                return (rgb, mask)
-            else:
-                # Return defaults if no file provided
-                return (default_image, default_mask)
-
-        except Exception as e:
-            print(f"Error loading EXR: {str(e)}")
-            # Return defaults on error
-            return (default_image, default_mask)
-
-NODE_CLASS_MAPPINGS = {
-    "ComfyUIDeployExternalEXR": ComfyUIDeployExternalEXR
-}
-
-NODE_DISPLAY_NAME_MAPPINGS = {
-    "ComfyUIDeployExternalEXR": "External EXR (ComfyUI Deploy)"
-}

comfy-nodes/external_exr_input.py

@@ -0,0 +1,93 @@
+import os
+os.environ["OPENCV_IO_ENABLE_OPENEXR"] = "1"
+import cv2 as cv
+import numpy as np
+import torch
+from folder_paths import get_annotated_filepath
+
+def linear_to_srgb(np_array):
+    """Converts a linear RGB numpy array to sRGB."""
+    less = np_array <= 0.0031308
+    np_array[less] = np_array[less] * 12.92
+    np_array[~less] = np.power(np_array[~less], 1/2.4) * 1.055 - 0.055
+    return np_array
+
+class ExternalExrInput:
+    """
+    Node to load a single EXR image from a local file path.
+    """
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "exr_file": ("STRING", {"default": "path/to/image.exr"}),
+                "tonemap": (["linear", "sRGB", "Reinhard"], {"default": "sRGB"}),
+            },
+            "optional": {
+                "default_image": ("IMAGE",),
+                "default_mask": ("MASK",),
+            }
+        }
+
+    RETURN_TYPES = ("IMAGE", "MASK")
+    RETURN_NAMES = ("image", "mask",)
+    FUNCTION = "run"
+    CATEGORY = "🔗ComfyDeploy/EXR"
+
+    def run(self, exr_file, tonemap, default_image=None, default_mask=None):
+        image = None
+        try:
+            if exr_file and exr_file.strip() != "":
+                exr_path = get_annotated_filepath(exr_file)
+                if os.path.exists(exr_path):
+                    image = cv.imread(exr_path, cv.IMREAD_UNCHANGED).astype(np.float32)
+                else:
+                    print(f"Warning: File not found at {exr_path}")
+            
+            if image is None:
+                raise ValueError("Image could not be loaded.")
+
+            if len(image.shape) == 2: # Grayscale
+                image = np.repeat(image[..., np.newaxis], 3, axis=2)
+            
+            rgb = np.flip(image[:, :, :3], 2).copy() # BGR to RGB
+            
+            # Apply tonemapping
+            if tonemap == "sRGB":
+                rgb = linear_to_srgb(rgb)
+                rgb = np.clip(rgb, 0, 1)
+            elif tonemap == "Reinhard":
+                rgb = np.clip(rgb, 0, None)
+                rgb = rgb / (rgb + 1)
+                rgb = linear_to_srgb(rgb)
+                rgb = np.clip(rgb, 0, 1)
+            
+            rgb_tensor = torch.from_numpy(rgb).unsqueeze(0)
+            
+            # Handle alpha/mask
+            if image.shape[2] > 3:
+                mask = np.clip(image[:, :, 3], 0, 1)
+            else:
+                mask = np.ones_like(rgb[:, :, 0])
+            mask_tensor = torch.from_numpy(mask).unsqueeze(0)
+
+            return (rgb_tensor, mask_tensor)
+
+        except Exception as e:
+            print(f"Error loading EXR file '{exr_file}': {e}")
+            if default_image is not None and default_mask is not None:
+                print("Returning default image.")
+                return (default_image, default_mask)
+            
+            print("Warning: Error loading EXR and no default image. Returning a black image.")
+            blank_image = torch.zeros((1, 64, 64, 3), dtype=torch.float32)
+            blank_mask = torch.zeros((1, 64, 64), dtype=torch.float32)
+            return (blank_image, blank_mask)
+
+NODE_CLASS_MAPPINGS = {
+    "ExternalExrInput": ExternalExrInput
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "ExternalExrInput": "External EXR Input (ComfyDeploy)"
+} 

comfy-nodes/external_exr_output.py

@@ -0,0 +1,73 @@
+import os
+os.environ["OPENCV_IO_ENABLE_OPENEXR"] = "1"
+import cv2 as cv
+import torch
+import numpy as np
+import folder_paths
+
+def srgb_to_linear(np_array):
+    """Converts an sRGB numpy array to linear RGB."""
+    less = np_array <= 0.0404482362771082
+    np_array[less] = np_array[less] / 12.92
+    np_array[~less] = np.power((np_array[~less] + 0.055) / 1.055, 2.4)
+    return np_array
+
+class ExternalExrOutput:
+    """
+    Node to save a single image as an EXR file to a local path.
+    """
+    def __init__(self):
+        self.output_dir = folder_paths.get_output_directory()
+        self.type = "output"
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "images": ("IMAGE",),
+                "filepath": ("STRING", {"default": "/tmp/output.exr"}),
+                "tonemap": (["linear", "sRGB"], {"default": "linear"}),
+            },
+        }
+
+    RETURN_TYPES = ()
+    FUNCTION = "run"
+    OUTPUT_NODE = True
+    CATEGORY = "🔗ComfyDeploy/EXR"
+
+    def run(self, images, filepath, tonemap):
+        if not filepath.endswith(".exr"):
+            raise ValueError("Filepath must end with '.exr'")
+
+        output_dir = os.path.dirname(filepath)
+        if not os.path.isabs(output_dir):
+            raise ValueError("Filepath must be an absolute path.")
+        
+        os.makedirs(output_dir, exist_ok=True)
+        
+        # We only process the first image in the batch
+        image_tensor = images[0]
+        
+        linear = image_tensor.cpu().numpy().astype(np.float32)
+        
+        # If the source is sRGB, convert to linear
+        if tonemap == "sRGB":
+            linear[...,:3] = srgb_to_linear(linear[...,:3])
+
+        # Convert RGB to BGR for OpenCV
+        bgr = np.flip(linear, 2).copy()
+        
+        # Save the image
+        cv.imwrite(filepath, bgr)
+        
+        print(f"Saved EXR file to: {filepath}")
+
+        return {"ui": {"images": [{"filename": os.path.basename(filepath), "subfolder": os.path.dirname(filepath), "type": self.type}]}}
+
+NODE_CLASS_MAPPINGS = {
+    "ExternalExrOutput": ExternalExrOutput
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "ExternalExrOutput": "External EXR Output (ComfyDeploy)"
+} 

comfy-nodes/external_exr_sequence_input.py

@@ -0,0 +1,159 @@
+import os
+os.environ["OPENCV_IO_ENABLE_OPENEXR"] = "1"
+import cv2 as cv
+import numpy as np
+import torch
+import re
+from folder_paths import get_annotated_filepath
+
+def linear_to_srgb(np_array):
+    """Converts a linear RGB numpy array to sRGB."""
+    less = np_array <= 0.0031308
+    np_array[less] = np_array[less] * 12.92
+    np_array[~less] = np.power(np_array[~less], 1/2.4) * 1.055 - 0.055
+    return np_array
+
+class ExternalExrSequenceInput:
+    """
+    Node to load a sequence of EXR images from a local filepath pattern, a directory,
+    or a single file within a sequence.
+    """
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "path_or_pattern": ("STRING", {"default": "path/to/frames_or_pattern"}),
+                "tonemap": (["linear", "sRGB", "Reinhard"], {"default": "sRGB"}),
+                "start_frame": ("INT", {"default": 1, "min": 1}),
+                "end_frame": ("INT", {"default": 50, "min": 1}),
+            },
+            "optional": {
+                "default_image": ("IMAGE",),
+                "default_mask": ("MASK",),
+            }
+        }
+
+    RETURN_TYPES = ("IMAGE", "MASK")
+    RETURN_NAMES = ("image", "mask",)
+    FUNCTION = "run"
+    CATEGORY = "🔗ComfyDeploy/EXR"
+
+    def get_image_paths(self, path_input, start_frame, end_frame):
+        image_paths = []
+        
+        # Case 1: Input is a C-style pattern
+        if '%' in path_input:
+            print(f"Pattern detected: {path_input}")
+            for i in range(start_frame, end_frame + 1):
+                fpath = get_annotated_filepath(path_input % i)
+                if os.path.exists(fpath):
+                    image_paths.append(fpath)
+            return image_paths
+
+        annotated_path = get_annotated_filepath(path_input)
+        
+        # Case 2: Input is a directory
+        if os.path.isdir(annotated_path):
+            print(f"Directory detected: {annotated_path}")
+            files_in_dir = sorted(os.listdir(annotated_path))
+            for filename in files_in_dir:
+                if not filename.lower().endswith('.exr'):
+                    continue
+                
+                matches = re.findall(r'\d+', filename)
+                if not matches:
+                    continue
+                
+                frame_number = int(matches[-1])
+                if start_frame <= frame_number <= end_frame:
+                    image_paths.append(os.path.join(annotated_path, filename))
+            return image_paths
+            
+        # Case 3: Input is a single file from a sequence
+        if os.path.isfile(annotated_path):
+            print(f"Single file detected: {annotated_path}. Attempting to find sequence.")
+            base_dir = os.path.dirname(annotated_path)
+            filename = os.path.basename(annotated_path)
+
+            matches = list(re.finditer(r'(\d+)', filename))
+            if not matches: # It's a single file with no frame number
+                return [annotated_path]
+
+            last_match = matches[-1]
+            num_start_pos, num_end_pos = last_match.span()
+            prefix = filename[:num_start_pos]
+            suffix = filename[num_end_pos:]
+            padding = len(last_match.group(0))
+
+            for i in range(start_frame, end_frame + 1):
+                potential_filename = f"{prefix}{str(i).zfill(padding)}{suffix}"
+                potential_path = os.path.join(base_dir, potential_filename)
+                if os.path.exists(potential_path):
+                    image_paths.append(potential_path)
+            return image_paths
+
+        return [] # Return empty if no cases match
+
+    def run(self, path_or_pattern, tonemap, start_frame, end_frame, default_image=None, default_mask=None):
+        try:
+            image_paths = self.get_image_paths(path_or_pattern, start_frame, end_frame)
+            if not image_paths:
+                raise ValueError(f"No EXR files found for '{path_or_pattern}' between frames {start_frame}-{end_frame}.")
+
+            print(f"Found {len(image_paths)} EXR files to load.")
+            rgb_frames = []
+            mask_frames = []
+            
+            for path in image_paths:
+                image = cv.imread(path, cv.IMREAD_UNCHANGED)
+                if image is None:
+                    print(f"Warning: Could not read file {path}, skipping.")
+                    continue
+                
+                image = image.astype(np.float32)
+                if len(image.shape) == 2:
+                    image = np.repeat(image[..., np.newaxis], 3, axis=2)
+                
+                rgb = np.flip(image[:, :, :3], 2).copy()
+                
+                if tonemap == "sRGB":
+                    rgb = linear_to_srgb(rgb)
+                    rgb = np.clip(rgb, 0, 1)
+                elif tonemap == "Reinhard":
+                    rgb = np.clip(rgb, 0, None)
+                    rgb = rgb / (rgb + 1)
+                    rgb = linear_to_srgb(rgb)
+                    rgb = np.clip(rgb, 0, 1)
+                
+                rgb_frames.append(torch.from_numpy(rgb))
+                
+                if image.shape[2] > 3:
+                    mask = np.clip(image[:, :, 3], 0, 1)
+                else:
+                    mask = np.ones_like(rgb[:, :, 0])
+                mask_frames.append(torch.from_numpy(mask))
+
+            if not rgb_frames:
+                raise ValueError("No frames were loaded successfully.")
+
+            print(f"Successfully loaded {len(rgb_frames)} frames into a batch.")
+            return (torch.stack(rgb_frames, 0), torch.stack(mask_frames, 0))
+
+        except Exception as e:
+            print(f"Error loading EXR sequence: {e}")
+            if default_image is not None and default_mask is not None:
+                print("Returning default image.")
+                return (default_image, default_mask)
+
+            print("Warning: Error loading sequence and no default image. Returning a black image.")
+            blank_image = torch.zeros((1, 64, 64, 3), dtype=torch.float32)
+            blank_mask = torch.zeros((1, 64, 64), dtype=torch.float32)
+            return (blank_image, blank_mask)
+
+NODE_CLASS_MAPPINGS = {
+    "ExternalExrSequenceInput": ExternalExrSequenceInput
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "ExternalExrSequenceInput": "External EXR Sequence Input (ComfyDeploy)"
+} 

comfy-nodes/external_exr_sequence_output.py

@@ -0,0 +1,88 @@
+import os
+os.environ["OPENCV_IO_ENABLE_OPENEXR"] = "1"
+import cv2 as cv
+import torch
+import numpy as np
+import re
+
+def srgb_to_linear(np_array):
+    """Converts an sRGB numpy array to linear RGB."""
+    less = np_array <= 0.0404482362771082
+    np_array[less] = np_array[less] / 12.92
+    np_array[~less] = np.power((np_array[~less] + 0.055) / 1.055, 2.4)
+    return np_array
+
+class ExternalExrSequenceOutput:
+    """
+    Node to save a sequence of images as EXR files to a local directory.
+    It uses a filepath pattern like 'path/to/frame_%04d.exr' to save each frame.
+    """
+    def __init__(self):
+        self.type = "output"
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "images": ("IMAGE",),
+                "filepath_pattern": ("STRING", {"default": "/tmp/exr_sequence/frame_%04d.exr"}),
+                "tonemap": (["linear", "sRGB"], {"default": "linear"}),
+            },
+        }
+
+    RETURN_TYPES = ()
+    FUNCTION = "run"
+    OUTPUT_NODE = True
+    CATEGORY = "🔗ComfyDeploy/EXR"
+
+    def run(self, images, filepath_pattern, tonemap):
+        # Basic validation for the filepath pattern
+        if not re.search(r'%0?\d+d', filepath_pattern):
+            raise ValueError("Filepath pattern must contain a C-style format specifier like '%04d'.")
+            
+        if not filepath_pattern.endswith(".exr"):
+            raise ValueError("Filepath pattern must end with '.exr'.")
+
+        output_dir = os.path.dirname(filepath_pattern)
+        if not os.path.isabs(output_dir):
+            raise ValueError("Filepath must be an absolute path.")
+            
+        os.makedirs(output_dir, exist_ok=True)
+        
+        # Convert tensor to numpy array
+        linear_images = images.cpu().numpy().astype(np.float32)
+        
+        # If the source is sRGB, convert to linear
+        if tonemap == "sRGB":
+            srgb_to_linear(linear_images[...,:3])
+        
+        # Convert RGB to BGR for OpenCV
+        bgr_images = np.flip(linear_images, 3).copy()
+
+        results = []
+        for i, bgr_image in enumerate(bgr_images):
+            frame_num = i + 1
+            try:
+                # Use the pattern to format the full file path
+                file_path = filepath_pattern % frame_num
+            except TypeError:
+                raise ValueError("Invalid format specifier in filepath_pattern. Use '%d', '%04d', etc.")
+
+            # Save the image
+            cv.imwrite(file_path, bgr_image)
+            
+            results.append({
+                "filename": os.path.basename(file_path),
+                "subfolder": os.path.dirname(file_path),
+                "type": self.type,
+            })
+        
+        return {"ui": {"images": results}}
+
+NODE_CLASS_MAPPINGS = {
+    "ExternalExrSequenceOutput": ExternalExrSequenceOutput
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "ExternalExrSequenceOutput": "External EXR Sequence Output (ComfyDeploy)"
+} 

comfy-nodes/http_exr_input.py

@@ -0,0 +1,111 @@
+import os
+os.environ["OPENCV_IO_ENABLE_OPENEXR"] = "1"
+import cv2 as cv
+import numpy as np
+import torch
+import requests
+
+def linear_to_srgb(np_array):
+    """Converts a linear RGB numpy array to sRGB."""
+    less = np_array <= 0.0031308
+    np_array[less] = np_array[less] * 12.92
+    np_array[~less] = np.power(np_array[~less], 1/2.4) * 1.055 - 0.055
+    return np_array
+
+class HttpExrInput:
+    """
+    Node to load a single EXR image from a URL, with optional tonemapping.
+    This node is designed to be used in a ComfyDeploy environment where input files are provided via signed URLs.
+    """
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "get_signed_url": ("STRING", {"multiline": True, "default": ""}),
+                "tonemap": (["linear", "sRGB", "Reinhard"], {"default": "sRGB"}),
+                "seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff}),
+            },
+            "optional": {
+                "default_image": ("IMAGE",),
+                "default_mask": ("MASK",),
+            },
+        }
+
+    RETURN_TYPES = ("IMAGE", "MASK")
+    RETURN_NAMES = ("image", "mask",)
+    FUNCTION = "run"
+    CATEGORY = "🔗ComfyDeploy/EXR"
+
+    def load_exr_from_data(self, exr_data):
+        try:
+            nparr = np.frombuffer(exr_data, np.uint8)
+            # Use cv.IMREAD_UNCHANGED to keep all channels (e.g., alpha)
+            image = cv.imdecode(nparr, cv.IMREAD_UNCHANGED)
+            if image is None:
+                raise ValueError("Failed to decode EXR data.")
+            return image.astype(np.float32)
+        except Exception as e:
+            print(f"Error decoding EXR data: {e}")
+            return None
+
+    def run(self, get_signed_url, tonemap, seed, default_image=None, default_mask=None):
+        if not get_signed_url or get_signed_url.strip() == "":
+            print("Warning: No input URL provided. Returning default image if available.")
+            if default_image is not None and default_mask is not None:
+                return (default_image, default_mask)
+            
+            print("Warning: No input URL and no default image. Returning a black image.")
+            blank_image = torch.zeros((1, 64, 64, 3), dtype=torch.float32)
+            blank_mask = torch.zeros((1, 64, 64), dtype=torch.float32)
+            return (blank_image, blank_mask)
+
+        image = None
+        try:
+            print(f"Fetching EXR from URL: {get_signed_url}")
+            response = requests.get(get_signed_url)
+            response.raise_for_status()
+            image = self.load_exr_from_data(response.content)
+        except requests.exceptions.RequestException as e:
+            print(f"Error fetching EXR from URL {get_signed_url}: {e}")
+
+        if image is None:
+            print("Warning: Could not load or decode EXR image. Returning default image if available.")
+            if default_image is not None and default_mask is not None:
+                return (default_image, default_mask)
+
+            print("Warning: Failed to load EXR and no default image. Returning a black image.")
+            blank_image = torch.zeros((1, 64, 64, 3), dtype=torch.float32)
+            blank_mask = torch.zeros((1, 64, 64), dtype=torch.float32)
+            return (blank_image, blank_mask)
+                    
+        # BGR to RGB conversion and channel handling
+        if len(image.shape) == 2: # Grayscale
+            image = np.repeat(image[..., np.newaxis], 3, axis=2)
+        
+        rgb = np.flip(image[:, :, :3], 2).copy() # OpenCV loads as BGR, convert to RGB
+        
+        # Tonemapping
+        if tonemap == "sRGB":
+            rgb = linear_to_srgb(rgb)
+            rgb = np.clip(rgb, 0, 1)
+        elif tonemap == "Reinhard":
+            rgb = np.clip(rgb, 0, None) # Ensure no negative values
+            rgb = rgb / (rgb + 1)
+            rgb = linear_to_srgb(rgb)
+            rgb = np.clip(rgb, 0, 1)
+        
+        # Handle alpha channel if it exists
+        if image.shape[2] > 3:
+            mask = np.clip(image[:, :, 3], 0, 1)
+        else:
+            mask = np.ones_like(rgb[:, :, 0]) # Create a full white mask if no alpha
+
+        return (torch.from_numpy(rgb).unsqueeze(0), torch.from_numpy(mask).unsqueeze(0),)
+
+NODE_CLASS_MAPPINGS = {
+    "HttpExrInput": HttpExrInput
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "HttpExrInput": "HTTP EXR Input (ComfyDeploy)"
+} 

comfy-nodes/http_exr_output.py

@@ -0,0 +1,80 @@
+import os
+os.environ["OPENCV_IO_ENABLE_OPENEXR"] = "1"
+import cv2 as cv
+import torch
+import numpy as np
+import requests
+
+def srgb_to_linear(np_array):
+    """Converts an sRGB numpy array to linear RGB."""
+    less = np_array <= 0.0404482362771082
+    np_array[less] = np_array[less] / 12.92
+    np_array[~less] = np.power((np_array[~less] + 0.055) / 1.055, 2.4)
+    return np_array
+
+class HttpExrOutput:
+    """
+    Node to save a single EXR image to a pre-signed URL.
+    This node is designed for ComfyDeploy to upload the generated EXR file.
+    """
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "images": ("IMAGE",),
+                "put_signed_url": ("STRING", {"multiline": True, "default": ""}),
+                "tonemap": (["linear", "sRGB"], {"default": "linear"}),
+            },
+            "hidden": {"prompt": "PROMPT", "extra_pnginfo": "EXTRA_PNGINFO"},
+        }
+
+    RETURN_TYPES = ()
+    FUNCTION = "run"
+    OUTPUT_NODE = True
+    CATEGORY = "ComfyDeploy/EXR"
+
+    def run(self, images, put_signed_url, tonemap, prompt=None, extra_pnginfo=None):
+        if not put_signed_url or put_signed_url.strip() == "":
+            print("Warning: No put_signed_url provided. Nothing will be uploaded.")
+            return {"ui": {"images": []}}
+        
+        # We process only the first image of the batch
+        image_tensor = images[0]
+        
+        # Convert tensor to numpy array, assuming it's in range [0, 1]
+        linear = image_tensor.cpu().numpy().astype(np.float32)
+        
+        # If the source is sRGB, convert to linear
+        if tonemap == "sRGB":
+            linear[...,:3] = srgb_to_linear(linear[...,:3])
+
+        # Convert RGB to BGR for OpenCV
+        bgr = np.flip(linear, 2).copy()
+        
+        results = []
+        try:
+            # Encode the image to the EXR format in memory
+            is_success, buffer = cv.imencode(".exr", bgr)
+            if not is_success:
+                raise Exception("Failed to encode image to EXR format.")
+            
+            # Upload the image data to the pre-signed URL
+            response = requests.put(put_signed_url, data=buffer.tobytes(), headers={'Content-Type': 'image/x-exr'})
+            response.raise_for_status()
+            
+            print(f"Successfully uploaded EXR to: {put_signed_url}")
+            # The UI can optionally display a link or confirmation
+            results.append({"url": put_signed_url, "output_id": "output_http_exr"})
+            
+        except Exception as e:
+            print(f"Error uploading EXR to signed URL: {e}")
+
+        return {"ui": {"images": results}}
+
+NODE_CLASS_MAPPINGS = {
+    "HttpExrOutput": HttpExrOutput
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "HttpExrOutput": "HTTP EXR Output (ComfyDeploy)"
+} 

comfy-nodes/http_exr_sequence_input.py

@@ -0,0 +1,126 @@
+import os
+os.environ["OPENCV_IO_ENABLE_OPENEXR"] = "1"
+import cv2 as cv
+import numpy as np
+import torch
+import requests
+import json
+
+def linear_to_srgb(np_array):
+    """Converts a linear RGB numpy array to sRGB."""
+    less = np_array <= 0.0031308
+    np_array[less] = np_array[less] * 12.92
+    np_array[~less] = np.power(np_array[~less], 1/2.4) * 1.055 - 0.055
+    return np_array
+
+class HttpExrSequenceInput:
+    """
+    Node to load a sequence of EXR images from a list of URLs provided as a JSON string.
+    """
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "urls_json": ("STRING", {"multiline": True, "default": "[]"}),
+                "tonemap": (["linear", "sRGB", "Reinhard"], {"default": "sRGB"}),
+                "seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff}),
+            },
+            "optional": {
+                "default_image": ("IMAGE",),
+                "default_mask": ("MASK",),
+            },
+        }
+
+    RETURN_TYPES = ("IMAGE", "MASK")
+    RETURN_NAMES = ("image", "mask",)
+    FUNCTION = "run"
+    CATEGORY = "🔗ComfyDeploy/EXR"
+
+    def load_exr_from_data(self, exr_data):
+        try:
+            nparr = np.frombuffer(exr_data, np.uint8)
+            image = cv.imdecode(nparr, cv.IMREAD_UNCHANGED)
+            if image is None:
+                raise ValueError("Failed to decode EXR data.")
+            return image.astype(np.float32)
+        except Exception as e:
+            print(f"Error decoding EXR data: {e}")
+            return None
+
+    def run(self, urls_json, tonemap, seed, default_image=None, default_mask=None):
+        try:
+            urls = json.loads(urls_json)
+            if not isinstance(urls, list) or not all(isinstance(u, str) for u in urls):
+                raise ValueError("urls_json must be a JSON array of URL strings.")
+        except (json.JSONDecodeError, ValueError) as e:
+            print(f"Error parsing urls_json: {e}. Using default image if available.")
+            urls = []
+
+        if not urls:
+            if default_image is not None and default_mask is not None:
+                return (default_image, default_mask)
+            
+            print("Warning: No valid URLs and no default image. Returning a black image.")
+            blank_image = torch.zeros((1, 64, 64, 3), dtype=torch.float32)
+            blank_mask = torch.zeros((1, 64, 64), dtype=torch.float32)
+            return (blank_image, blank_mask)
+
+        rgb_frames = []
+        mask_frames = []
+
+        for url in urls:
+            image = None
+            try:
+                print(f"Fetching EXR from URL: {url}")
+                response = requests.get(url)
+                response.raise_for_status()
+                image = self.load_exr_from_data(response.content)
+            except requests.exceptions.RequestException as e:
+                print(f"Error fetching EXR from URL {url}: {e}")
+            
+            if image is None:
+                print(f"Warning: Could not decode EXR from {url}. Skipping frame.")
+                continue
+                
+            if len(image.shape) == 2: # Grayscale
+                image = np.repeat(image[..., np.newaxis], 3, axis=2)
+            
+            rgb = np.flip(image[:, :, :3], 2).copy() # BGR to RGB
+            
+            if tonemap == "sRGB":
+                rgb = linear_to_srgb(rgb)
+                rgb = np.clip(rgb, 0, 1)
+            elif tonemap == "Reinhard":
+                rgb = np.clip(rgb, 0, None)
+                rgb = rgb / (rgb + 1)
+                rgb = linear_to_srgb(rgb)
+                rgb = np.clip(rgb, 0, 1)
+            
+            rgb_frames.append(torch.from_numpy(rgb))
+
+            if image.shape[2] > 3:
+                mask = np.clip(image[:, :, 3], 0, 1)
+            else:
+                mask = np.ones_like(rgb[:, :, 0])
+            mask_frames.append(torch.from_numpy(mask))
+
+        if not rgb_frames:
+            print("Could not load any frames. Returning default image if available.")
+            if default_image is not None and default_mask is not None:
+                return (default_image, default_mask)
+            
+            print("Warning: Failed to load any frames and no default image. Returning a black image.")
+            blank_image = torch.zeros((1, 64, 64, 3), dtype=torch.float32)
+            blank_mask = torch.zeros((1, 64, 64), dtype=torch.float32)
+            return (blank_image, blank_mask)
+
+        print(f"Loaded {len(rgb_frames)} frames successfully.")
+        return (torch.stack(rgb_frames, 0), torch.stack(mask_frames, 0))
+
+NODE_CLASS_MAPPINGS = {
+    "HttpExrSequenceInput": HttpExrSequenceInput
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "HttpExrSequenceInput": "HTTP EXR Sequence Input (ComfyDeploy)"
+} 

comfy-nodes/http_exr_sequence_output.py

@@ -0,0 +1,91 @@
+import os
+os.environ["OPENCV_IO_ENABLE_OPENEXR"] = "1"
+import cv2 as cv
+import torch
+import numpy as np
+import requests
+import json
+
+def srgb_to_linear(np_array):
+    """Converts an sRGB numpy array to linear RGB."""
+    less = np_array <= 0.0404482362771082
+    np_array[less] = np_array[less] / 12.92
+    np_array[~less] = np.power((np_array[~less] + 0.055) / 1.055, 2.4)
+    return np_array
+
+class HttpExrSequenceOutput:
+    """
+    Node to save a sequence of images as EXR files to a list of pre-signed URLs.
+    """
+    def __init__(self):
+        self.type = "output"
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "images": ("IMAGE",),
+                "upload_urls_json": ("STRING", {"multiline": True, "default": "[]"}),
+                "tonemap": (["linear", "sRGB"], {"default": "linear"}),
+            },
+        }
+
+    RETURN_TYPES = ()
+    FUNCTION = "run"
+    OUTPUT_NODE = True
+    CATEGORY = "🔗ComfyDeploy/EXR"
+
+    def run(self, images, upload_urls_json, tonemap):
+        try:
+            upload_urls = json.loads(upload_urls_json)
+            if not isinstance(upload_urls, list) or not all(isinstance(u, str) for u in upload_urls):
+                raise ValueError("upload_urls_json must be a JSON array of URL strings.")
+        except (json.JSONDecodeError, ValueError) as e:
+            print(f"Error parsing upload_urls_json: {e}. Aborting upload.")
+            return {"ui": {"images": []}}
+
+        if not upload_urls:
+            print("Warning: No upload URLs provided. Nothing will be uploaded.")
+            return {"ui": {"images": []}}
+        
+        if len(images) != len(upload_urls):
+            print(f"Warning: Mismatch between number of images ({len(images)}) and upload URLs ({len(upload_urls)}). Aborting upload.")
+            return {"ui": {"images": []}}
+
+        # Convert tensor to numpy array
+        linear_images = images.cpu().numpy().astype(np.float32)
+        
+        # If the source is sRGB, convert all images to linear
+        if tonemap == "sRGB":
+            srgb_to_linear(linear_images[...,:3])
+        
+        # Convert RGB to BGR for OpenCV
+        bgr_images = np.flip(linear_images, 3).copy()
+
+        results = []
+        for i, (bgr_image, url) in enumerate(zip(bgr_images, upload_urls)):
+            try:
+                # Encode the image to the EXR format in memory
+                is_success, buffer = cv.imencode(".exr", bgr_image)
+                if not is_success:
+                    raise Exception("Failed to encode image to EXR format.")
+                
+                # Upload the image data to the pre-signed URL
+                response = requests.put(url, data=buffer.tobytes(), headers={'Content-Type': 'image/x-exr'})
+                response.raise_for_status()
+                
+                print(f"Successfully uploaded frame {i+1} to: {url}")
+                results.append({"url": url})
+                
+            except Exception as e:
+                print(f"Error uploading frame {i+1} to {url}: {e}")
+
+        return {"ui": {"images": results}}
+
+NODE_CLASS_MAPPINGS = {
+    "HttpExrSequenceOutput": HttpExrSequenceOutput
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "HttpExrSequenceOutput": "HTTP EXR Sequence Output (ComfyDeploy)"
+} 

comfy-nodes/output_exr.py

@@ -1,78 +0,0 @@
-# In file: comfyui-deploy/comfy-nodes/output_exr.py
-
-import os
-import numpy as np
-import folder_paths
-
-# Try to set up OpenCV for EXR writing.
-try:
-    os.environ["OPENCV_IO_ENABLE_OPENEXR"] = "1"
-    import cv2
-    OPENCV_AVAILABLE = True
-except ImportError:
-    print("Warning: OpenCV not found for ComfyDeployOutputEXR. Please add opencv-python-headless to requirements.txt")
-    OPENCV_AVAILABLE = False
-
-# ALIGNED: Renamed class to match project conventions
-class ComfyDeployOutputEXR:
-    def __init__(self):
-        self.output_dir = folder_paths.get_output_directory()
-        self.type = "output"
-
-    @classmethod
-    def INPUT_TYPES(s):
-        return {
-            "required": {
-                "images": ("IMAGE", ),
-                "filename_prefix": ("STRING", {"default": "ComfyDeploy_EXR"})
-            },
-            # ADDED: Optional output_id for consistency with other ComfyDeploy nodes
-            "optional": {
-                "output_id": ("STRING", {"multiline": False, "default": "output_exr"}),
-            },
-        }
-    
-    RETURN_TYPES = ()
-    # ALIGNED: Changed function name to 'run'
-    FUNCTION = "run"
-    OUTPUT_NODE = True
-    # ALIGNED: Matched the category name
-    CATEGORY = "🔗ComfyDeploy"
-    DESCRIPTION = "Saves the input images as EXR (HDR) files."
-
-    def run(self, images, filename_prefix="ComfyDeploy_EXR", output_id="output_exr"):
-        if not OPENCV_AVAILABLE:
-            raise ImportError("OpenCV is required to save EXR files. Please ensure opencv-python-headless is in requirements.txt.")
-
-        full_output_folder, filename, counter, subfolder, filename_prefix = (
-            folder_paths.get_save_image_path(
-                filename_prefix, self.output_dir, images[0].shape[1], images[0].shape[0]
-            )
-        )
-        results = list()
-
-        for image in images:
-            image_np = image.cpu().numpy()
-
-            if image_np.dtype != np.float32:
-                 image_np = image_np.astype(np.float32)
-
-            file = f"{filename}_{counter:05}.exr"
-            file_path = os.path.join(full_output_folder, file)
-
-            image_np_bgr = cv2.cvtColor(image_np, cv2.COLOR_RGB2BGR)
-            cv2.imwrite(file_path, image_np_bgr)
-
-            results.append({
-                "filename": file,
-                "subfolder": subfolder,
-                "type": self.type,
-                "output_id": output_id, # ADDED
-            })
-            counter += 1
-
-        return {"ui": {"images": results}}
-
-# ALIGNED: Mappings are defined at the bottom of the node file in this project
-NODE_CLASS_MAPPINGS = {"ComfyDeployOutputEXR": ComfyDeployOutputEXR}
-NODE_DISPLAY_NAME_MAPPINGS = {"ComfyDeployOutputEXR": "EXR Output (ComfyDeploy)"}

globals.py

@@ -18,7 +18,6 @@ class Status(Enum):
     SUCCESS = "success"
     FAILED = "failed"
     UPLOADING = "uploading"
-    CANCELLED = "cancelled"
 
 
 class StreamingPrompt(BaseModel):
@@ -57,11 +56,27 @@ streaming_prompt_metadata: dict[str, StreamingPrompt] = {}
 class BinaryEventTypes:
     PREVIEW_IMAGE = 1
     UNENCODED_PREVIEW_IMAGE = 2
+    EXR_IMAGE = 4
 
 
 max_output_id_length = 24
 
 
+async def send_exr(image_data, sid=None, output_id: str = None):
+    max_length = max_output_id_length
+    output_id = output_id[:max_length]
+    padded_output_id = output_id.ljust(max_length, "\x00")
+    encoded_output_id = padded_output_id.encode("ascii", "replace")
+
+    bytesIO = BytesIO()
+    # 10 bytes for the output_id
+    bytesIO.write(encoded_output_id)
+    bytesIO.write(image_data)
+
+    preview_bytes = bytesIO.getvalue()
+    await send_bytes(BinaryEventTypes.EXR_IMAGE, preview_bytes, sid=sid)
+
+
 async def send_image(image_data, sid=None, output_id: str = None):
     max_length = max_output_id_length
     output_id = output_id[:max_length]

pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "comfyui-deploy"
 description = "Open source comfyui deployment platform, a vercel for generative workflow infra."
-version = "2.3.6"
+version = "2.1.0"
 license = { file = "LICENSE" }
 dependencies = ["aiofiles", "pydantic", "opencv-python", "imageio-ffmpeg", "tabulate", "brotli"]
 

web-plugin/snapshot-utils.js

@@ -1,68 +0,0 @@
-// Snapshot Utilities
-// Centralized snapshot fetching with ComfyUI version fallback
-
-/**
- * Fetches the current snapshot with ComfyUI version fallback
- * If the snapshot response has null comfyui field, it will fetch the latest ComfyUI version
- * and update the snapshot with the comfyui_hash
- *
- * @param {Function} getDataFn - Function that returns { apiKey, apiUrl } for ComfyUI version API calls
- * @returns {Promise<Object>} - The snapshot data with comfyui field populated
- */
-export async function fetchSnapshot(getDataFn = null) {
-  try {
-    // Fetch the current snapshot
-    const response = await fetch("/snapshot/get_current");
-    if (!response.ok) {
-      throw new Error(`Snapshot fetch failed: ${response.status}`);
-    }
-
-    const snapshot = await response.json();
-
-    // Check if comfyui field is null and we have getDataFn for fallback
-    if (snapshot.comfyui === null && getDataFn) {
-      console.log(
-        "ComfyUI version is null in snapshot, fetching latest version..."
-      );
-
-      try {
-        const data = getDataFn();
-        if (data && data.apiKey) {
-          const comfyuiVersionResponse = await fetch(
-            `/comfyui-deploy/comfyui-version?api_url=${encodeURIComponent(
-              data.apiUrl || "https://api.comfydeploy.com"
-            )}`,
-            {
-              headers: {
-                Authorization: `Bearer ${data.apiKey}`,
-              },
-            }
-          );
-
-          if (comfyuiVersionResponse.ok) {
-            const versionData = await comfyuiVersionResponse.json();
-            if (versionData.comfyui_hash) {
-              console.log(
-                `Using ComfyUI hash from API: ${versionData.comfyui_hash}`
-              );
-              snapshot.comfyui = versionData.comfyui_hash;
-            }
-          } else {
-            console.warn(
-              "Failed to fetch ComfyUI version from API:",
-              comfyuiVersionResponse.status
-            );
-          }
-        }
-      } catch (error) {
-        console.warn("Error fetching ComfyUI version fallback:", error);
-        // Continue with original snapshot even if fallback fails
-      }
-    }
-
-    return snapshot;
-  } catch (error) {
-    console.error("Error fetching snapshot:", error);
-    throw error;
-  }
-}