Update app.py

app.py

@@ -231,6 +231,9 @@ def chat_interface(user_input, history, web_search, decoding_strategy, temperatu
     # Ensure the tokenizer is accessible within the function scope
     global tokenizer
     
+    # Ensure the input is correctly formatted as a dictionary
+    user_input = {"text": user_input}
+    
     # Perform model inference
     response = model_inference(
         user_prompt=user_input,
@@ -243,15 +246,18 @@ def chat_interface(user_input, history, web_search, decoding_strategy, temperatu
         tokenizer=tokenizer  # Pass tokenizer to the model_inference function
     )
     
-    # Return the response
-    return response
+    # Update chat history
+    history.append([user_input["text"], response])
+    
+    # Return the updated history
+    return history, response
 
 # Define the Gradio interface components
 interface = gr.Interface(
     fn=chat_interface,
     inputs=[
         gr.Textbox(label="User Input", placeholder="Type your message here..."),
-        gr.Textbox(label="Chat History", placeholder="Chat history will appear here..."),
+        gr.State([], label="Chat History"),
         gr.Checkbox(label="Perform Web Search"),
         gr.Radio(["Greedy", "Top P Sampling"], label="Decoding strategy"),
         gr.Slider(minimum=0.0, maximum=2.0, step=0.05, label="Sampling temperature", value=0.5),
@@ -259,8 +265,11 @@ interface = gr.Interface(
         gr.Slider(minimum=0.01, maximum=5.0, step=0.01, label="Repetition penalty", value=1),
         gr.Slider(minimum=0.01, maximum=0.99, step=0.01, label="Top P", value=0.9)
     ],
-    outputs=gr.Textbox(label="Assistant Response"),
-    live=True
+    outputs=[
+        gr.Chatbot(label="Assistant Response"),
+        gr.State([])
+    ],
+    live=False
 )
 
 # Launch the Gradio interface