Update app.py

app.py

@@ -3,24 +3,24 @@ import openai
 import fitz  # PyMuPDF for PDF processing
 import base64
 
-# Variable to store API key
+# Store API Key
 api_key = ""
 
-# Function to update API key
+# Function to update API Key
 def set_api_key(key):
     global api_key
     api_key = key
     return "API Key Set Successfully!"
 
-# Function to interact with OpenAI API
+# Function to interact with OpenAI API using conversation history
 def query_openai(messages, temperature, top_p, max_output_tokens):
     if not api_key:
-        return "Please enter your OpenAI API key first."
+        return ["Please enter your OpenAI API key first."], messages
 
     try:
-        openai.api_key = api_key  # Set API key dynamically
+        openai.api_key = api_key  # Set API Key dynamically
 
-        # Ensure numeric values for OpenAI parameters
+        # Ensure valid values for OpenAI parameters
         temperature = float(temperature) if temperature else 1.0
         top_p = float(top_p) if top_p else 1.0
         max_output_tokens = int(max_output_tokens) if max_output_tokens else 2048
@@ -32,35 +32,37 @@ def query_openai(messages, temperature, top_p, max_output_tokens):
             top_p=top_p,
             max_tokens=max_output_tokens
         )
-        return response["choices"][0]["message"]["content"]
+
+        bot_response = response["choices"][0]["message"]["content"]
+        messages.append({"role": "assistant", "content": bot_response})  # Store bot response in history
+        return messages, messages  # Return updated conversation
+
     except Exception as e:
-        return f"Error: {str(e)}"
+        return [f"Error: {str(e)}"], messages
 
-# Function to process image URL input
-def image_url_chat(image_url, text_query, temperature, top_p, max_output_tokens):
+# Image URL Chat
+def image_url_chat(image_url, text_query, messages, temperature, top_p, max_output_tokens):
     if not image_url or not text_query:
-        return "Please provide an image URL and a query."
-
-    messages = [
-        {"role": "user", "content": [
-            {"type": "image_url", "image_url": {"url": image_url}},  # Corrected format
-            {"type": "text", "text": text_query}
-        ]}
-    ]
+        return ["Please provide an image URL and a query."], messages
+
+    messages.append({"role": "user", "content": [
+        {"type": "image_url", "image_url": {"url": image_url}},
+        {"type": "text", "text": text_query}
+    ]})
     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):
+# Text Chat
+def text_chat(text_query, messages, temperature, top_p, max_output_tokens):
     if not text_query:
-        return "Please enter a query."
+        return ["Please enter a query."], messages
 
-    messages = [{"role": "user", "content": [{"type": "text", "text": text_query}]}]
+    messages.append({"role": "user", "content": text_query})
     return query_openai(messages, 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):
+# Image Chat
+def image_chat(image_file, text_query, messages, temperature, top_p, max_output_tokens):
     if image_file is None or not text_query:
-        return "Please upload an image and provide a query."
+        return ["Please upload an image and provide a query."], messages
 
     # Encode image as base64
     with open(image_file, "rb") as img:
@@ -68,39 +70,34 @@ def image_chat(image_file, text_query, temperature, top_p, max_output_tokens):
 
     image_data = f"data:image/jpeg;base64,{base64_image}"
 
-    messages = [
-        {"role": "user", "content": [
-            {"type": "image_url", "image_url": {"url": image_data}},  # Fixed format
-            {"type": "text", "text": text_query}
-        ]}
-    ]
+    messages.append({"role": "user", "content": [
+        {"type": "image_url", "image_url": {"url": image_data}},
+        {"type": "text", "text": text_query}
+    ]})
     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):
+# PDF Chat
+def pdf_chat(pdf_file, text_query, messages, temperature, top_p, max_output_tokens):
     if pdf_file is None or not text_query:
-        return "Please upload a PDF and provide a query."
+        return ["Please upload a PDF and provide a query."], messages
 
-    # Extract text from the first few pages
     doc = fitz.open(pdf_file)
-    text = "\n".join([page.get_text("text") for page in doc][:5])  # Limit extraction for performance
-
-    messages = [
-        {"role": "user", "content": [
-            {"type": "text", "text": text},  # Fixed format
-            {"type": "text", "text": text_query}
-        ]}
-    ]
+    text = "\n".join([page.get_text("text") for page in doc][:5])  # Extract text from first 5 pages
+
+    messages.append({"role": "user", "content": [
+        {"type": "text", "text": text},
+        {"type": "text", "text": text_query}
+    ]})
     return query_openai(messages, temperature, top_p, max_output_tokens)
 
-# Function to clear the chat (Fix: Returns the correct number of outputs)
+# Function to clear chat history and reset parameters
 def clear_chat():
-    return "", "", "", "", "", "", "", None, "", None, "", 1.0, 1.0, 2048
+    return [], [], [], [], [], [], [], None, "", None, "", 1.0, 1.0, 2048
 
 # Gradio UI Layout
 with gr.Blocks() as demo:
-    gr.Markdown("## GPT-4.5 Preview Chatbot")
-    
+    gr.Markdown("## GPT-4.5 Preview Conversational Chatbot")
+
     # API Key Input
     with gr.Row():
         api_key_input = gr.Textbox(label="Enter OpenAI API Key", type="password")
@@ -110,46 +107,53 @@ with gr.Blocks() as demo:
     with gr.Row():
         temperature = gr.Slider(0, 2, value=1.0, step=0.1, label="Temperature")
         top_p = gr.Slider(0, 1, value=1.0, step=0.1, label="Top-P")
-        max_output_tokens = gr.Slider(0, 16384, value=2048, step=512, label="Max Output Tokens")  # Changed default to 2048
-    
+        max_output_tokens = gr.Slider(0, 16384, value=2048, step=512, label="Max Output Tokens")
+
     with gr.Tabs():
         with gr.Tab("Image URL Chat"):
             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_output = gr.Chatbot(label="Conversation History", elem_id="chatbot1")
             image_url_button = gr.Button("Ask")
-        
+
         with gr.Tab("Text Chat"):
             text_query = gr.Textbox(label="Enter your query")
-            text_output = gr.Textbox(label="Response", interactive=False)
+            text_output = gr.Chatbot(label="Conversation History", elem_id="chatbot2")
             text_button = gr.Button("Ask")
-        
+
         with gr.Tab("Image Chat"):
             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_output = gr.Chatbot(label="Conversation History", elem_id="chatbot3")
             image_button = gr.Button("Ask")
-        
+
         with gr.Tab("PDF Chat"):
             pdf_upload = gr.File(label="Upload a PDF", type="filepath")
             pdf_text_query = gr.Textbox(label="Ask about the uploaded PDF")
-            pdf_output = gr.Textbox(label="Response", interactive=False)
+            pdf_output = gr.Chatbot(label="Conversation History", elem_id="chatbot4")
             pdf_button = gr.Button("Ask")
 
     # Clear chat button
     clear_button = gr.Button("Clear Chat")
 
+    # Chat Histories
+    image_url_chat_history = gr.State([])
+    text_chat_history = gr.State([])
+    image_chat_history = gr.State([])
+    pdf_chat_history = gr.State([])
+
     # 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)
-    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)
-    pdf_button.click(pdf_chat, [pdf_upload, pdf_text_query, temperature, top_p, max_output_tokens], pdf_output)
+    image_url_button.click(image_url_chat, [image_url, image_query, image_url_chat_history, temperature, top_p, max_output_tokens], [image_url_output, image_url_chat_history])
+    text_button.click(text_chat, [text_query, text_chat_history, temperature, top_p, max_output_tokens], [text_output, text_chat_history])
+    image_button.click(image_chat, [image_upload, image_text_query, image_chat_history, temperature, top_p, max_output_tokens], [image_output, image_chat_history])
+    pdf_button.click(pdf_chat, [pdf_upload, pdf_text_query, pdf_chat_history, temperature, top_p, max_output_tokens], [pdf_output, pdf_chat_history])
 
     # Fix: Clear button resets all necessary fields correctly
     clear_button.click(
         clear_chat,
         outputs=[
+            image_url_chat_history, text_chat_history, image_chat_history, pdf_chat_history,
             image_url, image_query, image_url_output, 
             text_query, text_output, 
             image_text_query, image_output,