Update app.py

app.py

@@ -81,7 +81,14 @@ def count_tokens(text: str) -> int:
     return len(tokenizer.encode(text))
 
 def truncate_history(history: list[tuple[str, str]], system_message: str, max_length: int) -> list[tuple[str, str]]:
-    """Truncates the conversation history to fit within the maximum token limit."""
+    """Truncates the conversation history to fit within the maximum token limit.
+    Args:
+        history: The conversation history (list of user/assistant tuples).
+        system_message: The system message.
+        max_length: The maximum number of tokens allowed.
+    Returns:
+        The truncated history.
+    """
     truncated_history = []
     system_message_tokens = count_tokens(system_message)
     current_length = system_message_tokens
@@ -93,55 +100,53 @@ def truncate_history(history: list[tuple[str, str]], system_message: str, max_le
         turn_tokens = user_tokens + assistant_tokens
 
         if current_length + turn_tokens <= max_length:
-            truncated_history.insert(0, (user_msg, assistant_msg))
+            truncated_history.insert(0, (user_msg, assistant_msg))  # Add to the beginning
             current_length += turn_tokens
         else:
             break  # Stop adding turns if we exceed the limit
 
     return truncated_history
 
-def respond(message, history: list[tuple[str, str]], system_message, max_tokens, temperature, top_p, clear_memory):
-    """Responds to a user message, maintaining conversation history using a streaming generator."""
-    # Check for the clear memory command (or if the Clear Memory button was triggered)
-    if message.lower() == "clear memory" or clear_memory:
-        return "", []  # Reset the chat
+def respond(
+    message,
+    history: list[tuple[str, str]],
+    system_message,  # System message is now an argument
+    max_tokens,
+    temperature,
+    top_p,
+):
+    """Responds to a user message, maintaining conversation history, using special tokens and message list."""
 
-    formatted_system_message = system_message  # Use the provided system message
-    truncated_history = truncate_history(history, formatted_system_message, MAX_CONTEXT_LENGTH - max_tokens - 100)
+    if message.lower() == "clear memory": # Check for the clear memory command
+        return "", []  # Return empty message and empty history to reset the chat
 
-    messages = [{"role": "system", "content": formatted_system_message}]
+    formatted_system_message = system_message  # Use the system_message argument
+    truncated_history = truncate_history(history, formatted_system_message, MAX_CONTEXT_LENGTH - max_tokens - 100) # Reserve space for the new message and some generation
+
+    messages = [{"role": "system", "content": formatted_system_message}] # Start with system message as before
     for user_msg, assistant_msg in truncated_history:
         if user_msg:
-            messages.append({"role": "user", "content": f"<|user|>\n{user_msg}</s>"})
+            messages.append({"role": "user", "content": f"<|user|>\n{user_msg}</s>"}) # Format history user message
         if assistant_msg:
-            messages.append({"role": "assistant", "content": f"<|assistant|>\n{assistant_msg}</s>"})
-    messages.append({"role": "user", "content": f"<|user|>\n{message}</s>"})
+            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
 
     response = ""
-    yielded_any = False
     try:
-        for chunk in client.chat_completion(
-            messages,
-            max_tokens=max_tokens,
-            stream=True,
-            temperature=temperature,
-            top_p=top_p,
-        ):
-            token = chunk.choices[0].delta.content
-            response += token
-            yield response
-            yielded_any = True
-    except StopAsyncIteration:
-        # If the generator stops without yielding any tokens, yield the accumulated response
-        if not yielded_any:
-            yield response
-        return
+      for chunk in client.chat_completion(
+          messages, # Send the messages list again, but with formatted content
+          max_tokens=max_tokens,
+          stream=True,
+          temperature=temperature,
+          top_p=top_p,
+      ):
+          token = chunk.choices[0].delta.content
+          response += token
+          yield response
     except Exception as e:
-        print(f"An error occurred: {e}")
-        yield "I'm sorry, I encountered an error. Please try again."
-    # Ensure at least one yield if no tokens were yielded inside the loop
-    if not yielded_any:
-        yield response
+      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."
 
 # --- Gradio Interface ---
 demo = gr.ChatInterface(
@@ -151,15 +156,19 @@ demo = gr.ChatInterface(
             value=default_nvc_prompt_template,
             label="System message",
             visible=True,
-            lines=10  # Increased height for easier reading of the prompt
+            lines=10,  # Increased height for more space to read the prompt
         ),
         gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
-        gr.Button("Clear Memory"),
+        gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.95,
+            step=0.05,
+            label="Top-p (nucleus sampling)",
+        ),
     ],
-    chatbot_kwargs={"format": "messages"}  # Use messages format to avoid deprecation warnings
 )
 
 if __name__ == "__main__":
-    demo.launch(share=True)  # Note: share=True is not supported on Hugging Face Spaces.
+    demo.launch(share=True)