s

app.py

@@ -1,68 +1,170 @@
+# Requires gradio==3.16.2
+# Requires huggingface_hub==0.11.0
+
 import gradio as gr
 from huggingface_hub import InferenceClient
+import logging
+from functools import lru_cache
+import html
+import signal
+
+# Setup logging
+logging.basicConfig(level=logging.INFO)
 
+# Initialize the Hugging Face Inference Client
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
+# Constants
+MAX_HISTORY_LENGTH = 5  # Adjust as needed
+
+# Graceful shutdown handler
+def shutdown_handler(signum, frame):
+    logging.info("Shutting down gracefully...")
+    exit(0)
+
+signal.signal(signal.SIGINT, shutdown_handler)
 
-def respond(
-    message: str,
-    history: list[tuple[str, str]],
-    system_message: str,
-    max_tokens: int,
-    temperature: float,
-    top_p: float,
-):
+def system_message_selector(choice, custom_message):
+    """
+    Selects the system message based on the user's choice or custom input.
+
+    Parameters:
+        choice (str): The persona choice selected by the user.
+        custom_message (str): A custom persona or system message provided by the user.
+
+    Returns:
+        str: The system message to be used in the conversation.
+    """
+    if custom_message:
+        return custom_message
+    elif choice == "Friendly Chatbot":
+        return "You are a friendly and helpful chatbot."
+    elif choice == "Professional Assistant":
+        return "You are a highly knowledgeable and professional assistant."
+    elif choice == "Curious Researcher":
+        return "You are a curious researcher who loves to explore new ideas."
+    else:
+        return "You are a helpful assistant."
+
+@lru_cache(maxsize=32)
+def get_response_from_model(messages_tuple, max_tokens, temperature, top_p):
+    """
+    Calls the Hugging Face Inference API to get a response.
+
+    Parameters:
+        messages_tuple (tuple): A tuple of messages to be sent to the model.
+        max_tokens (int): Maximum number of tokens for the response.
+        temperature (float): Sampling temperature.
+        top_p (float): Top-p (nucleus) sampling parameter.
+
+    Returns:
+        str: The generated response from the model.
+    """
+    # Convert tuple back to list of dicts
+    messages = [dict(m) for m in messages_tuple]
+    response = ""
+    for message in client.chat_completion(
+        messages,
+        max_tokens=max_tokens,
+        stream=True,
+        temperature=temperature,
+        top_p=top_p,
+    ):
+        token = message.choices[0].delta.content
+        response += token
+    return response
+
+def sanitize_input(text):
+    """
+    Sanitizes user input to prevent code injection or XSS attacks.
+
+    Parameters:
+        text (str): The user input text.
+
+    Returns:
+        str: The sanitized text.
+    """
+    return html.escape(text)
+
+def validate_parameters(max_tokens, temperature, top_p):
+    """
+    Validates input parameters.
+
+    Parameters:
+        max_tokens (int): Maximum number of tokens for the response.
+        temperature (float): Sampling temperature.
+        top_p (float): Top-p (nucleus) sampling parameter.
+
+    Returns:
+        tuple: (bool, str) indicating validity and an error message if invalid.
+    """
+    if not (1 <= max_tokens <= 2048):
+        return False, "Error: 'Max new tokens' must be between 1 and 2048."
+    if not (0.1 <= temperature <= 4.0):
+        return False, "Error: 'Temperature' must be between 0.1 and 4.0."
+    if not (0.1 <= top_p <= 1.0):
+        return False, "Error: 'Top-p' must be between 0.1 and 1.0."
+    return True, ""
+
+def respond(message, history, persona_choice, custom_persona, max_tokens, temperature, top_p):
     """
     Generates a response using the Hugging Face Inference API.
 
-    Args:
+    Parameters:
         message (str): User's current input.
         history (list[tuple[str, str]]): Previous conversation history.
-        system_message (str): Instructions for the model (e.g., persona details).
+        persona_choice (str): The selected persona.
+        custom_persona (str): Custom persona or system message.
         max_tokens (int): Maximum tokens allowed for the response.
-        temperature (float): Sampling temperature for randomness in the output.
-        top_p (float): Top-p (nucleus) sampling parameter.
+        temperature (float): Sampling temperature.
+        top_p (float): Top-p (nucleus sampling) parameter.
 
     Yields:
         str: The generated chatbot response.
     """
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+    # Validate parameters
+    is_valid, error_message = validate_parameters(max_tokens, temperature, top_p)
+    if not is_valid:
+        yield error_message
+        return
 
-    messages.append({"role": "user", "content": message})
+    # Sanitize user input
+    safe_message = sanitize_input(message)
+    safe_history = [(sanitize_input(u), sanitize_input(b)) for u, b in history]
 
-    response = ""
+    # Limit the history to the most recent exchanges
+    truncated_history = safe_history[-MAX_HISTORY_LENGTH:]
 
-    try:
-        for message in client.chat_completion(
-            messages,
-            max_tokens=max_tokens,
-            stream=True,
-            temperature=temperature,
-            top_p=top_p,
-        ):
-            token = message.choices[0].delta.content
-            response += token
-            yield response
-    except Exception as e:
-        yield f"Error: {str(e)}"
+    # Select system message
+    system_message = system_message_selector(persona_choice, custom_persona)
 
+    # Build messages with truncated history
+    messages = [{"role": "system", "content": system_message}]
+    for user_msg, bot_msg in truncated_history:
+        if user_msg:
+            messages.append({"role": "user", "content": user_msg})
+        if bot_msg:
+            messages.append({"role": "assistant", "content": bot_msg})
+    messages.append({"role": "user", "content": safe_message})
 
-def system_message_selector(choice):
-    if choice == "Friendly Chatbot":
-        return "You are a friendly and helpful chatbot."
-    elif choice == "Professional Assistant":
-        return "You are a highly knowledgeable and professional assistant."
-    elif choice == "Curious Researcher":
-        return "You are a curious researcher who loves to explore new ideas."
-    else:
-        return "You are a helpful assistant."
+    # Log the request
+    logging.info(f"Received message: {safe_message}")
 
+    try:
+        # Convert messages to a tuple of tuples for caching
+        messages_tuple = tuple(tuple(m.items()) for m in messages)
+
+        # Use caching to optimize performance
+        response = get_response_from_model(
+            messages_tuple,
+            max_tokens,
+            temperature,
+            top_p,
+        )
+        yield response
+    except Exception as e:
+        logging.error(f"An error occurred: {e}")
+        yield "I'm sorry, but something went wrong. Please try again."
 
 # Create the UI components
 system_message_radio = gr.Radio(
@@ -71,23 +173,37 @@ system_message_radio = gr.Radio(
     label="Choose a Persona",
 )
 
-# ChatInterface with dynamic system message selection
+system_message_textbox = gr.Textbox(
+    placeholder="Enter custom persona or system message...",
+    label="Custom Persona (Optional)",
+)
+
+max_tokens_slider = gr.Slider(
+    minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"
+)
+
+temperature_slider = gr.Slider(
+    minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"
+)
+
+top_p_slider = gr.Slider(
+    minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"
+)
+
+# Create the ChatInterface directly with additional inputs
 demo = gr.ChatInterface(
-    respond,
+    fn=respond,
     additional_inputs=[
         system_message_radio,
-        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)",
-        ),
+        system_message_textbox,
+        max_tokens_slider,
+        temperature_slider,
+        top_p_slider,
     ],
+    allow_reset_history=True,
+    title="Customizable Chatbot Interface",
+    description="Choose a persona or enter a custom one, and adjust parameters as needed.",
 )
 
-
 if __name__ == "__main__":
     demo.launch()