Update app.py

app.py

@@ -1,6 +1,6 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
-from transformers import AutoTokenizer  # Import the tokenizer
+from transformers import AutoTokenizer
 
 # Use the appropriate tokenizer for your model.
 tokenizer = AutoTokenizer.from_pretrained("HuggingFaceH4/zephyr-7b-beta")
@@ -134,26 +134,25 @@ def respond(
     temperature,
     top_p,
 ):
-    """Responds to a user message, maintaining conversation history, using special tokens and message list."""
+    """Responds to a user message, maintaining conversation history.  Returns history."""
 
-    formatted_system_message = nvc_prompt_template
+    formatted_system_message = system_message # Use the provided system message
 
-    truncated_history = truncate_history(history, formatted_system_message, MAX_CONTEXT_LENGTH - max_tokens - 100) # Reserve space for the new message and some generation
+    truncated_history = truncate_history(history, formatted_system_message, MAX_CONTEXT_LENGTH - max_tokens - 100)
 
-    messages = [{"role": "system", "content": formatted_system_message}] # Start with system message as before
+    messages = [{"role": "system", "content": formatted_system_message}]
     for user_msg, assistant_msg in truncated_history:
         if user_msg:
-            messages.append({"role": "user", "content": f"<|user|>\n{user_msg}</s>"}) # Format history user message
+            messages.append({"role": "user", "content": f"<|user|>\n{user_msg}</s>"})
         if assistant_msg:
-            messages.append({"role": "assistant", "content": f"<|assistant|>\n{assistant_msg}</s>"}) # Format history assistant message
-
-    messages.append({"role": "user", "content": f"<|user|>\n{message}</s>"}) # Format current user message
+            messages.append({"role": "assistant", "content": f"<|assistant|>\n{assistant_msg}</s>"})
 
+    messages.append({"role": "user", "content": f"<|user|>\n{message}</s>"})
 
     response = ""
     try:
         for chunk in client.chat_completion(
-            messages, # Send the messages list again, but with formatted content
+            messages,
             max_tokens=max_tokens,
             stream=True,
             temperature=temperature,
@@ -161,10 +160,14 @@ def respond(
         ):
             token = chunk.choices[0].delta.content
             response += token
-            yield response
+            # Crucial:  Yield the *updated* history.  Append the *current* response.
+            yield history + [(message, response)]  # Build the (user, bot) tuple.
+
     except Exception as e:
-        print(f"An error occurred: {e}") # It's a good practice add a try-except block
-        yield "I'm sorry, I encountered an error. Please try again."
+        print(f"An error occurred: {e}")
+        error_message = "I'm sorry, I encountered an error. Please try again."
+        # Also yield the history, even in case of error, to preserve context.
+        yield history + [(message, error_message)]
 
 
 def clear_memory(history, chat_history):
@@ -176,11 +179,14 @@ def clear_memory(history, chat_history):
 
 
 # --- Gradio Interface ---
-with gr.Blocks() as demo:  # Use gr.Blocks for more control
-    chatbot = gr.Chatbot(label="Roos NVC Chatbot")  # Chatbot UI element
-    msg = gr.Textbox(label="Your Message")      # Text input for the user
-
-    with gr.Accordion("Settings", open=False): # Settings section, initially hidden
+with gr.Blocks() as demo:
+    chatbot = gr.Chatbot(label="Roos NVC Chatbot")
+    msg = gr.Textbox(label="Your Message")
+    with gr.Row():  # Put the Send button and Clear button on the same row
+        send_btn = gr.Button("Send")  # Add a Send button
+        clear_btn = gr.Button("Clear Memory")
+
+    with gr.Accordion("Settings", open=False):
       system_message = gr.Textbox(value=nvc_prompt_template, label="System message")
       max_tokens = gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens")
       temperature = gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature")
@@ -191,11 +197,12 @@ with gr.Blocks() as demo:  # Use gr.Blocks for more control
                 step=0.05,
                 label="Top-p (nucleus sampling)",
             )
-    clear_btn = gr.Button("Clear Memory") #Clear memory button
 
-    msg.submit(respond, [msg, chatbot, system_message, max_tokens, temperature, top_p], chatbot)
-    clear_btn.click(clear_memory, [chatbot, chatbot], [msg, chatbot])
 
+    # Connect both Enter key *and* Send button to the respond function
+    msg.submit(respond, [msg, chatbot, system_message, max_tokens, temperature, top_p], chatbot)
+    send_btn.click(respond, [msg, chatbot, system_message, max_tokens, temperature, top_p], chatbot)
+    clear_btn.click(clear_memory, [chatbot, chatbot], [msg, chatbot])  #Corrected
 
 if __name__ == "__main__":
     demo.launch()