src/streamlit_app.py

import streamlit as st
import pandas as pd
import joblib
import numpy as np
from sklearn.neighbors import NearestNeighbors
from sklearn.preprocessing import MinMaxScaler 
import os 

# --- Load Assets (Model and Data) ---
# Uses caching to load heavy files (model, idlist, scaler, and large DataFrame) only once.
@st.cache_resource
def load_assets():
    # File Names
    MODEL_FILENAME = 'src/book_recommendation_model.jobjob'
    IDLIST_FILENAME = 'src/knn_idlist.jobjob'
    SCALER_FILENAME = 'src/min_max_scaler.jobjob'
    DATA_FILENAME = 'src/books.csv'

    try:
        # 1. Load Saved Objects
        model = joblib.load(MODEL_FILENAME)
        idlist = joblib.load(IDLIST_FILENAME)
        scaler = joblib.load(SCALER_FILENAME)

        # 2. Load the DataFrame
        df = pd.read_csv(DATA_FILENAME, on_bad_lines='skip')
        df.columns = df.columns.str.strip() # Clean column names
        
        # 3. Check for essential column
        if 'title' not in df.columns:
            st.error("The DataFrame must contain a 'title' column.")
            return None, None, None, None 

        # Successful return of 4 objects
        return model, df, idlist, scaler
    
    except FileNotFoundError:
        st.error(f"Required files not found. Ensure '{MODEL_FILENAME}', '{IDLIST_FILENAME}', '{SCALER_FILENAME}', and '{DATA_FILENAME}' are in the same directory.")
        return None, None, None, None

# Load assets upon application start
loaded_model, df, idlist, scaler = load_assets()

# --- Recommendation Function ---
def get_recommendations(book_name, df_data, id_list_data):
    """
    Retrieves the titles of books similar to the given book name.
    """
    # Find the index of the input book
    try:
        book_idx = df_data[df_data['title'] == book_name].index[0]
    except IndexError:
        return ["Error: The specified book title was not found in the dataset."], False

    # Get neighbor indices from the pre-calculated idlist
    # Note: idlist[book_idx] contains the indices of the neighbors
    neighbor_indices = id_list_data[book_idx]

    # Retrieve the titles of the neighboring books
    # .unique() helps handle potential duplicates in the dataset
    recommended_titles = df_data.loc[neighbor_indices, 'title'].unique().tolist()
    
    # Remove the queried book itself from the list of recommendations
    if book_name in recommended_titles:
        recommended_titles.remove(book_name)
    
    # Return the top 5 (or N-1 if N neighbors were used) recommendations
    return recommended_titles[:5], True 

# --- Streamlit Interface ---

if loaded_model is not None and df is not None:
    st.title("📚 Book Recommendation System")
    st.markdown("This application uses a pre-trained K-Nearest Neighbors model to suggest similar books.")

    # Select box options
    book_options = df['title'].unique()
    
    selected_book = st.selectbox(
        "Please select a book:", 
        options=book_options,
        index=0
    )

    st.markdown("---")

    if st.button("Generate Recommendations"):
        with st.spinner('Searching for similar books...'):
            recommendations, success = get_recommendations(selected_book, df, idlist)

            if success:
                st.subheader(f"Recommendations Similar to '{selected_book}':")
                
                # Display recommendations as a numbered list
                for i, title in enumerate(recommendations, 1):
                    st.success(f"{i}. **{title}**")
            else:
                st.error(recommendations[0])