Update app.py

app.py

@@ -2,8 +2,6 @@ import gradio as gr
 import openai
 import fitz  # PyMuPDF for PDF processing
 import base64
-from io import BytesIO
-from PIL import Image
 
 # Variable to store API key
 api_key = ""
@@ -41,7 +39,7 @@ def query_openai(messages, temperature, top_p, max_output_tokens):
 # Function to process image URL input
 def image_url_chat(image_url, text_query, temperature, top_p, max_output_tokens):
     if not image_url or not text_query:
-        return "Please provide an image URL and a query.", None
+        return "Please provide an image URL and a query."
 
     messages = [
         {"role": "user", "content": [
@@ -49,8 +47,7 @@ def image_url_chat(image_url, text_query, temperature, top_p, max_output_tokens)
             {"type": "text", "text": text_query}
         ]}
     ]
-    response = query_openai(messages, temperature, top_p, max_output_tokens)
-    return response, image_url  # Return image URL to display it
+    return query_openai(messages, temperature, top_p, max_output_tokens)
 
 # Function to process text input
 def text_chat(text_query, temperature, top_p, max_output_tokens):
@@ -63,7 +60,7 @@ def text_chat(text_query, temperature, top_p, max_output_tokens):
 # Function to process uploaded image input
 def image_chat(image_file, text_query, temperature, top_p, max_output_tokens):
     if image_file is None or not text_query:
-        return "Please upload an image and provide a query.", None
+        return "Please upload an image and provide a query."
 
     # Encode image as base64
     with open(image_file, "rb") as img:
@@ -77,11 +74,7 @@ def image_chat(image_file, text_query, temperature, top_p, max_output_tokens):
             {"type": "text", "text": text_query}
         ]}
     ]
-    response = query_openai(messages, temperature, top_p, max_output_tokens)
-    
-    # Convert base64 image to PIL Image for display in Gradio
-    img = Image.open(BytesIO(base64.b64decode(base64_image)))
-    return response, img  # Return image for Gradio to display
+    return query_openai(messages, temperature, top_p, max_output_tokens)
 
 # Function to process uploaded PDF input
 def pdf_chat(pdf_file, text_query, temperature, top_p, max_output_tokens):
@@ -124,7 +117,6 @@ with gr.Blocks() as demo:
             image_url = gr.Textbox(label="Enter Image URL")
             image_query = gr.Textbox(label="Ask about the Image")
             image_url_output = gr.Textbox(label="Response", interactive=False)
-            image_url_display = gr.Image(label="Image", interactive=False)
             image_url_button = gr.Button("Ask")
         
         with gr.Tab("Text Chat"):
@@ -136,7 +128,6 @@ with gr.Blocks() as demo:
             image_upload = gr.File(label="Upload an Image", type="filepath")
             image_text_query = gr.Textbox(label="Ask about the uploaded image")
             image_output = gr.Textbox(label="Response", interactive=False)
-            image_display = gr.Image(label="Uploaded Image", interactive=False)
             image_button = gr.Button("Ask")
         
         with gr.Tab("PDF Chat"):
@@ -150,18 +141,18 @@ with gr.Blocks() as demo:
 
     # Button Click Actions
     api_key_button.click(set_api_key, inputs=[api_key_input], outputs=[api_key_output])
-    image_url_button.click(image_url_chat, [image_url, image_query, temperature, top_p, max_output_tokens], [image_url_output, image_url_display])
+    image_url_button.click(image_url_chat, [image_url, image_query, temperature, top_p, max_output_tokens], image_url_output)
     text_button.click(text_chat, [text_query, temperature, top_p, max_output_tokens], text_output)
-    image_button.click(image_chat, [image_upload, image_text_query, temperature, top_p, max_output_tokens], [image_output, image_display])
+    image_button.click(image_chat, [image_upload, image_text_query, temperature, top_p, max_output_tokens], image_output)
     pdf_button.click(pdf_chat, [pdf_upload, pdf_text_query, temperature, top_p, max_output_tokens], pdf_output)
 
     # Fix: Clear button resets all necessary fields correctly
     clear_button.click(
         clear_chat,
         outputs=[
-            image_url, image_query, image_url_output, image_url_display, 
+            image_url, image_query, image_url_output, 
             text_query, text_output, 
-            image_text_query, image_output, image_display, 
+            image_text_query, image_output, 
             pdf_upload, pdf_text_query, pdf_output, 
             temperature, top_p, max_output_tokens
         ]