Update app.py

app.py

@@ -10,27 +10,34 @@ ACCESS_TOKEN = os.getenv("HF_TOKEN")
 print("Access token loaded.")
 
 # Function to encode image to base64
-def encode_image(image):
-    if image is None:
+def encode_image(image_path):
+    if not image_path:
+        print("No image path provided")
         return None
     
-    # Convert to PIL Image if needed
-    if not isinstance(image, Image.Image):
-        try:
-            image = Image.open(image)
-        except Exception as e:
-            print(f"Error opening image: {e}")
-            return None
-    
-    # Convert to RGB if image has an alpha channel (RGBA)
-    if image.mode == 'RGBA':
-        image = image.convert('RGB')
-    
-    # Encode to base64
-    buffered = io.BytesIO()
-    image.save(buffered, format="JPEG")
-    img_str = base64.b64encode(buffered.getvalue()).decode("utf-8")
-    return img_str
+    try:
+        print(f"Encoding image from path: {image_path}")
+        
+        # If it's already a PIL Image
+        if isinstance(image_path, Image.Image):
+            image = image_path
+        else:
+            # Try to open the image file
+            image = Image.open(image_path)
+        
+        # Convert to RGB if image has an alpha channel (RGBA)
+        if image.mode == 'RGBA':
+            image = image.convert('RGB')
+        
+        # Encode to base64
+        buffered = io.BytesIO()
+        image.save(buffered, format="JPEG")
+        img_str = base64.b64encode(buffered.getvalue()).decode("utf-8")
+        print("Image encoded successfully")
+        return img_str
+    except Exception as e:
+        print(f"Error encoding image: {e}")
+        return None
 
 def respond(
     message,
@@ -221,20 +228,18 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
     )
     print("Chatbot interface created.")
     
-    with gr.Row():
-        # Multimodal textbox for messages (combines text and file uploads)
-        msg = gr.MultimodalTextbox(
-            placeholder="Type a message or upload images...",
-            show_label=False,
-            container=False,
-            scale=12,
-            file_types=["image"],
-            file_count="multiple",
-            sources=["upload"]
-        )
-
-    # Send button for messages
-    submit_btn = gr.Button("Send", variant="primary")
+    # Multimodal textbox for messages (combines text and file uploads)
+    msg = gr.MultimodalTextbox(
+        placeholder="Type a message or upload images...",
+        show_label=False,
+        container=False,
+        scale=12,
+        file_types=["image"],
+        file_count="multiple",
+        sources=["upload"]
+    )
+    
+    # Note: We're removing the separate submit button since MultimodalTextbox has its own
 
     # Create accordion for settings
     with gr.Accordion("Settings", open=False):
@@ -396,54 +401,79 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
 
     # Function for the chat interface
     def user(user_message, history):
+        # Debug logging for troubleshooting
+        print(f"User message received: {user_message}")
+        
         # Skip if message is empty (no text and no files)
-        if (not user_message["text"] or user_message["text"].strip() == "") and not user_message["files"]:
+        if not user_message or (not user_message.get("text") and not user_message.get("files")):
+            print("Empty message, skipping")
             return history
         
-        # Process images and text into a display message
-        display_message = ""
+        # Prepare multimodal message format
+        text_content = user_message.get("text", "").strip()
+        files = user_message.get("files", [])
         
-        # Add text if present
-        if user_message["text"] and user_message["text"].strip() != "":
-            display_message += user_message["text"]
+        print(f"Text content: {text_content}")
+        print(f"Files: {files}")
         
-        # Add image references if present
-        file_displays = []
-        for file in user_message["files"]:
-            file_displays.append(file)
+        # If both text and files are empty, skip
+        if not text_content and not files:
+            print("No content to display")
+            return history
         
-        # Return updated history with display message
-        if file_displays:
-            return history + [[(display_message, file_displays), None]]
+        # Process multimodal content
+        if files:
+            # For multimodal messages with files
+            for file_path in files:
+                print(f"Processing file: {file_path}")
+                if not file_path:
+                    continue
+                
+                # Add a combined message with text and file
+                history.append([(text_content, file_path), None])
+                # Reset text content for subsequent files if there are multiple
+                text_content = ""
+            
+            return history
         else:
-            return history + [[display_message, None]]
+            # For text-only messages
+            history.append([text_content, None])
+            return history
     
-    # Define chat interface
+    # Define bot response function
     def bot(history, system_msg, max_tokens, temperature, top_p, freq_penalty, seed, provider, api_key, custom_model, search_term, selected_model):
-        # Extract the last user message
+        # Check if history is valid
         if not history or len(history) == 0:
+            print("No history to process")
             return history
         
+        # Extract the last user message
         user_message = history[-1][0]
+        print(f"Processing user message: {user_message}")
         
-        # Determine if the message is multimodal or text-only
-        is_multimodal = False
-        text_parts = ""
+        # Determine if the message contains images
+        text_content = ""
         image_files = []
         
-        # Process text and images from the message
         if isinstance(user_message, tuple):
-            text_parts = user_message[0]
-            image_files = user_message[1]
-            is_multimodal = True
+            # Tuple format: (text, image_path)
+            text_content = user_message[0] if user_message[0] else ""
+            # Handle both single image path and list of paths
+            if isinstance(user_message[1], list):
+                image_files = user_message[1]
+            else:
+                image_files = [user_message[1]]
+            print(f"Multimodal message detected - Text: {text_content}, Images: {image_files}")
         else:
-            text_parts = user_message
+            # Text-only message
+            text_content = user_message
+            print(f"Text-only message detected: {text_content}")
         
         # Process message through respond function
         history[-1][1] = ""
         for response in respond(
-            text_parts,               # Text part
-            image_files if is_multimodal else None,  # Image part
+            text_content,
+            image_files if image_files else None,
             history[:-1],
             system_msg,
             max_tokens,
@@ -460,7 +490,7 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
             history[-1][1] = response
             yield history
 
-    # Event handlers
+    # Event handlers - only using the MultimodalTextbox's built-in submit functionality
     msg.submit(
         user,
         [msg, chatbot],
@@ -478,23 +508,6 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
         [msg]
     )
     
-    submit_btn.click(
-        user,
-        [msg, chatbot],
-        [chatbot],
-        queue=False
-    ).then(
-        bot,
-        [chatbot, system_message_box, max_tokens_slider, temperature_slider, top_p_slider, 
-         frequency_penalty_slider, seed_slider, provider_radio, byok_textbox, custom_model_box, 
-         model_search_box, featured_model_radio],
-        [chatbot]
-    ).then(
-        lambda: {"text": "", "files": []},  # Clear inputs after submission
-        None,
-        [msg]
-    )
-    
     # Connect the model filter to update the radio choices
     model_search_box.change(
         fn=filter_models,