app.py

import os
import pandas as pd
import streamlit as st
import logging
from transformers import pipeline

# Configure logging
logging.basicConfig(level=logging.DEBUG)

# Load Hugging Face model
MODEL_NAME = "mistralai/Mistral-Large-Instruct-2407"  # Hugging Face Model name
llm_pipeline = pipeline("text-generation", model=MODEL_NAME)

# Load datasets
def load_datasets():
    """Load datasets from CSV files."""
    try:
        electronics_df = pd.read_csv('electronics.csv')
        fashion_df = pd.read_csv('fashion.csv')
        return electronics_df, fashion_df
    except Exception as e:
        logging.error(f"Error loading datasets: {e}")
        st.error("Error loading datasets.")
        st.stop()

electronics_df, fashion_df = load_datasets()

# Keywords for routing queries
electronics_keywords = [
    'electronics', 'device', 'gadget', 'battery', 'performance', 
    'phone', 'mobile', 'laptop', 'tv', 'bluetooth', 'speakers', 
    'washing machine', 'headphones', 'camera', 'tablet', 'charger', 
    'smartwatch', 'refrigerator'
]
fashion_keywords = [
    'fashion', 'clothing', 'size', 'fit', 'material', 'shirt', 
    'pants', 'coats', 'shoes', 'girls dress', 'sarees', 'skirts', 
    'jackets', 'sweaters', 'suits', 'accessories', 't-shirts'
]

# LLM-based function
def query_llm(prompt):
    """Query the LLM for responses."""
    try:
        responses = llm_pipeline(prompt, max_length=150, num_return_sequences=1)
        return responses[0]['generated_text'].strip()
    except Exception as e:
        logging.error(f"Error querying the LLM: {e}")
        return "Sorry, I'm having trouble processing your request right now."

def determine_category(query):
    """Determine the category based on the query."""
    query_lower = query.lower()
    if any(keyword in query_lower for keyword in electronics_keywords):
        logging.debug(f"Query '{query}' categorized as 'electronics'.")
        return 'electronics'
    elif any(keyword in query_lower for keyword in fashion_keywords):
        logging.debug(f"Query '{query}' categorized as 'fashion'.")
        return 'fashion'
    else:
        logging.debug(f"Query '{query}' categorized as 'general'.")
        return 'general'

# Fetch response based on query
def get_response(user_input):
    """Determine the category and fetch the appropriate response."""
    if 'hi' in user_input.lower() or 'hello' in user_input.lower():
        return "Hi, welcome to the customer support chatbot. How can I help you?"

    category = determine_category(user_input)
    
    if category == 'electronics':
        response = electronics_response(user_input)
    elif category == 'fashion':
        response = fashion_response(user_input)
    else:
        # Use LLM for more complex queries
        response = query_llm(user_input)

    return response

# Streamlit Interface remains the same
def main():
    st.title("Customer Support Chatbot")

    if 'chat_history' not in st.session_state:
        st.session_state.chat_history = []

    user_input = st.text_input("Type your message here:")

    if st.button("Send"):
        if user_input:
            response_message = get_response(user_input)
            st.session_state.chat_history.append({"role": "user", "content": user_input})
            st.session_state.chat_history.append({"role": "assistant", "content": response_message})
    
    for message in st.session_state.chat_history:
        st.markdown(f"{message['role'].capitalize()}: {message['content']}")

if __name__ == "__main__":
    main()