Update app_dialogue.py

app_dialogue.py

@@ -59,9 +59,16 @@ logger = logging.getLogger()
 SEED = 38
 set_seed(38)
 
-def convert_to_rgb(image):
-    # `image.convert("RGB")` would only work for .jpg images, as it creates a wrong background
-    # for transparent images. The call to `alpha_composite` handles this case
+import requests 
+from io import BytesIO
+import uuid
+
+from PIL import Image
+
+from PIL import Image
+import tempfile
+
+def convert_to_rgb_pil(image):
     if image.mode == "RGB":
         return image
 
@@ -69,27 +76,44 @@ def convert_to_rgb(image):
     background = Image.new("RGBA", image_rgba.size, (255, 255, 255))
     alpha_composite = Image.alpha_composite(background, image_rgba)
     alpha_composite = alpha_composite.convert("RGB")
-    return alpha_composite
 
+    # Save the converted image to a temporary file
+    temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".jpg")
+    temp_file_path = temp_file.name
+    alpha_composite.save(temp_file_path)
+    temp_file.close()
+    return temp_file_path  # Return the path to the saved image
+    
 
-# Conversion between PIL Image <-> base64 <-> Markdown utils
-def pil_to_base64(pil_image):
-    """
-    Convert an PIL image into base64 string representation
-    """
-    buffered = BytesIO()
-    pil_image.save(buffered, format="JPEG")  # You can change the format as per your image type
-    encoded_image = base64.b64encode(buffered.getvalue()).decode("utf-8")
-    return encoded_image
-
-
-def pil_to_markdown_im(image):
-    """
-    Convert a PIL image into markdown filled with the base64 string representation.
-    """
-    img_b64_str = pil_to_base64(image)
-    img_str = f'<img src="data:image/png;base64,{img_b64_str}" />'
-    return img_str
+def convert_to_rgb(filepath_or_pilimg):
+    # `image.convert("RGB")` would only work for .jpg images, as it creates a wrong background
+    # for transparent images. The call to `alpha_composite` handles this case
+    if isinstance(filepath_or_pilimg, PIL.Image.Image):
+        return convert_to_rgb_pil(filepath_or_pilimg)
+    with Image.open(filepath_or_pilimg) as image:
+        # Check if the image is already in the RGB format
+        if image.mode == "RGB":
+            return filepath_or_pilimg  # If already in RGB, return the original path
+
+        # Convert image to RGBA
+        image_rgba = image.convert("RGBA")
+        
+        # Create a white background image of the same size
+        background = Image.new("RGBA", image_rgba.size, (255, 255, 255))
+        
+        # Composite the original image over the white background
+        alpha_composite = Image.alpha_composite(background, image_rgba)
+        
+        # Convert the composited image to RGB format
+        alpha_composite = alpha_composite.convert("RGB")
+
+        # Save the converted image to a temporary file
+        temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".jpg")
+        temp_file_path = temp_file.name
+        alpha_composite.save(temp_file_path)
+        temp_file.close()
+
+        return temp_file_path  # Return the path to the saved image
 
 
 def base64_to_pil(encoded_image):
@@ -159,17 +183,33 @@ def isolate_images_urls(prompt_list):
     ]
     ```
     """
+    print(f"******* isolate_images_urls *******")
+    print(f"params: prompt_list is - {prompt_list}")
+    
     linearized_list = []
     for prompt in prompt_list:
+        print(f"inside FOR loop: prompt in prompt_list is - {prompt}")
         # Prompt can be either a string, or a PIL image
         if isinstance(prompt, PIL.Image.Image):
+            print(f"inside first IF in FOR loop: prompt is of type PIL.Image.Image")
+            linearized_list.append(prompt)
+            print(f"linearized_list after append is - {linearized_list}")
+        elif isinstance(prompt, str) and "/tmp/gradio/" in prompt: #isinstance(prompt, PIL.Image.Image):
+            print(f"inside IF in FOR loop: prompt is a string and is a path for temporary file")
             linearized_list.append(prompt)
-        elif isinstance(prompt, str):
+            print(f"linearized_list after append is - {linearized_list}")
+        elif isinstance(prompt, str) and "/tmp/gradio/" not in prompt:
+            print(f"inside ELIF in FOR loop: prompt is a string and is NOT a path for temporary file")
             if "<fake_token_around_image>" not in prompt:
+                print(f"inside IF inside ELIF in FOR loop: '<fake_token_around_image>' is NOT in prompt")
                 linearized_list.append(prompt)
+                print(f"linearized_list after append is - {linearized_list}")
             else:
+                print(f"inside ELSE inside ELIF in FOR loop: '<fake_token_around_image>' IS IN prompt")
                 prompt_splitted = prompt.split("<fake_token_around_image>")
+                print(f"prompt_splitted is - {prompt_splitted}")
                 for ps in prompt_splitted:
+                    print(f"Inside FOR loop inside FOR loop: ps in prompt_split is {ps}")
                     if ps == "":
                         continue
                     if ps.startswith("<image:"):
@@ -181,6 +221,7 @@ def isolate_images_urls(prompt_list):
                 f"Unrecognized type for `prompt`. Got {type(type(prompt))}. Was expecting something in [`str`,"
                 " `PIL.Image.Image`]"
             )
+            print(f"linearized_list to be returned is - {linearized_list}")
     return linearized_list
 
 
@@ -209,17 +250,27 @@ def user_prompt_list_to_markdown(user_prompt_list: List[Union[str, PIL.Image.Ima
     Convert a user prompt in the list format (i.e. elements are either a PIL image or a string) into
     the markdown format that is used for the chatbot history and rendering.
     """
+    print("********** user_prompt_list_to_markdown *********")
+    print(f" param : user_prompt_list is - {user_prompt_list}")
     resulting_string = ""
     for elem in user_prompt_list:
-        if isinstance(elem, str):
+        print(f"inside user_prompt_list_to_markdown, for loop on user_prompt_list")
+        print(f"elem is - {elem} ")
+        if isinstance(elem, str) and "/tmp/gradio/" not in elem:
             resulting_string += elem
-        elif isinstance(elem, PIL.Image.Image):
-            resulting_string += pil_to_markdown_im(convert_to_rgb(elem))
+            print(f"inside IF - when elem is string and is not temp image filepath. resulting_string is - {resulting_string}")
+        #elif isinstance(elem, str) and "/tmp/gradio/" in elem:
+        #    resulting_string += f"![](/file={convert_to_rgb(elem)})"  #f"![](/file={image})" 
+        #    print(f"inside first ELIF - when elem is string and is the temp image filepath. resulting_string is - {resulting_string}")
+        elif isinstance(elem, PIL.Image.Image) or "/tmp/gradio/" in elem:  #and "/tmp/gradio/" in elem:
+            resulting_string += f"![](/file={convert_to_rgb(elem)})" #pil_to_markdown_im(convert_to_rgb(elem))
+            print(f"inside the ELIF - when elem is an instance of  PIL.Image.Image or has /temp/file/path. resulting_string after pil_to_markdown_im() operation is - {resulting_string}")
         else:
             raise ValueError(
                 "Unknown type for `user_prompt_list`. Expected an element of type `str` or `PIL.Image.Image` and got"
                 f" `{type(elem)}`"
             )
+    print(f" final resulting_string that will be returned is - {resulting_string}")
     return resulting_string