app.py

import gradio as gr

def chat(user_input, history=[]):
  """
  Chat function that takes user input and conversation history,
  returns the model response and updates the history. Handles potential
  memory issues by clearing conversation history after a certain length.

  Args:
      user_input: String containing the user's message.
      history: List of tuples containing conversation history 
                (user_message, model_response).

  Returns:
      A tuple containing the model response and updated history (limited length).
  """
  # Update history with user input
  history.append((user_input, None))

  # Clear conversation history if it exceeds a certain length (adjust as needed)
  if len(history) > 10:
    history = history[-5:]  # Keep the most recent 5 interactions

  # Access the loaded model (replace with appropriate error handling)
  model = gr.get("chatbot_model")

  # Generate response using the model (consider error handling and retries)
  response = model(user_input, max_length=50, do_sample=True)[0]['generated_text']

  # Update history with model response
  history.append((None, response))

  return response, history

# Attempt to load the model from Hugging Face (consider error handling)
try:
  chatbot_model = gr.load("models/lucas-w/mental-health-chatbot-3")
except Exception as e:
  print(f"Error loading model: {e}")
  chatbot_model = None  # Handle the case where model loading fails

# Launch the Gradio interface with error handling
if chatbot_model is not None:
  interface = gr.Interface(
    fn=chat,
    inputs="textbox",
    outputs="textbox",
    interpretation="chat",
    title="Mental Health Chatbot",
    description="Talk to a mental health assistant )",
    elem_id="chat-container",
    css="""
      #chat-container {
        height: 400px;
        overflow-y: scroll;
      }
    """
  )

  interface.launch()
else:
  print("Failed to launch chatbot. Please check model availability and error messages.")