import logging
from transformers import pipeline, Conversation
import random

logger = logging.getLogger(__name__)
chatbot_pipeline = None
conversation_history = {}

def load_chatbot_model():
    global chatbot_pipeline
    try:
        logger.info("Loading DialoGPT chatbot model...")
        chatbot_pipeline = pipeline(
            "conversational",
            model="microsoft/DialoGPT-medium",
            device="cpu"  # Use "cuda" if GPU available
        )
        logger.info("✓ Chatbot model loaded successfully")
    except Exception as e:
        logger.error(f"✗ Failed to load chatbot model: {str(e)}")
        chatbot_pipeline = None

async def get_chatbot_response(user_text: str, user_id: str = "default") -> str:
    """
    Generate chatbot response using DialoGPT.
    Maintains conversation history per user.
    """
    global chatbot_pipeline, conversation_history
    
    try:
        if chatbot_pipeline is None:
            load_chatbot_model()
            if chatbot_pipeline is None:
                return get_fallback_response(user_text)
        
        logger.info(f"Chatbot: Processing '{user_text}'")
        
        # Initialize conversation for this user if needed
        if user_id not in conversation_history:
            conversation_history[user_id] = Conversation()
        
        # Add user input to conversation
        conversation = conversation_history[user_id]
        conversation.add_user_input(user_text)
        
        # Generate response
        response = chatbot_pipeline(conversation)
        bot_response = response.generated_responses[-1].strip()
        
        if not bot_response:
            bot_response = get_fallback_response(user_text)
        
        logger.info(f"✓ Chatbot Response: '{bot_response}'")
        return bot_response
        
    except Exception as e:
        logger.error(f"✗ Chatbot Error: {str(e)}")
        return get_fallback_response(user_text)

def get_fallback_response(user_text: str) -> str:
    """Fallback responses when model fails"""
    responses = [
        f"I understand: '{user_text}'. How can I assist?",
        f"Interesting point about '{user_text}'. Tell me more?",
        f"Regarding '{user_text}', what would you like to know?",
        "I'm listening. Please continue.",
        f"That's a great question about '{user_text}'!"
    ]
    return random.choice(responses)