import streamlit as st
import pandas as pd
import numpy as np
import tensorflow as tf
import joblib
import os
import zipfile
import tempfile

# Define static file paths
BASE_DIR = os.path.dirname(__file__)
ZIP_MODEL_PATH = os.path.join(BASE_DIR, "recommender_model.zip")
MOVIES_PATH = os.path.join(BASE_DIR, "movies.csv")
ENCODINGS_PATH = os.path.join(BASE_DIR, "encodings.pkl")

@st.cache_resource
def load_model():
    try:
        # Define extraction directory in a writable temp location
        extract_dir = os.path.join(tempfile.gettempdir(), "recommender_model_extracted")

        # Only extract if not already done
        if not os.path.exists(extract_dir):
            with zipfile.ZipFile(ZIP_MODEL_PATH, "r") as zip_ref:
                zip_ref.extractall(extract_dir)

        # Load model from extracted directory
        return tf.keras.models.load_model(extract_dir)

    except Exception as e:
        st.error(f"❌ Failed to load model:\n\n{e}")
        st.stop()

@st.cache_data
def load_assets():
    try:
        df_movies = pd.read_csv(MOVIES_PATH)
    except FileNotFoundError:
        st.error("❌ movies.csv not found.")
        st.stop()

    try:
        user_map, movie_map = joblib.load(ENCODINGS_PATH)
    except FileNotFoundError:
        st.error("❌ encodings.pkl not found.")
        st.stop()

    return df_movies, user_map, movie_map

# Load model and assets
model = load_model()
movies_df, user2idx, movie2idx = load_assets()
reverse_movie_map = {v: k for k, v in movie2idx.items()}

# App UI
st.title("🎬 TensorFlow Movie Recommender")
st.write("Select some movies you've liked to get personalized recommendations:")

# Movie selection UI
movie_titles = movies_df.set_index("movieId")["title"].to_dict()
movie_choices = [movie_titles[mid] for mid in movie2idx if mid in movie_titles]
selected_titles = st.multiselect("🎞️ Liked movies", sorted(movie_choices))

# Create ratings dictionary
user_ratings = {}
for title in selected_titles:
    movie_id = next((k for k, v in movie_titles.items() if v == title), None)
    if movie_id:
        user_ratings[movie_id] = 5.0

# Generate recommendations
if st.button("🎯 Get Recommendations"):
    if not user_ratings:
        st.warning("Please select at least one movie.")
    else:
        liked_indices = [movie2idx[m] for m in user_ratings if m in movie2idx]
        if not liked_indices:
            st.error("⚠️ No valid movie encodings found.")
            st.stop()

        # Get embedding averages and scores
        avg_embedding = tf.reduce_mean(model.layers[2](tf.constant(liked_indices)), axis=0, keepdims=True)
        all_movie_indices = tf.range(len(movie2idx))
        movie_embeddings = model.layers[3](all_movie_indices)
        scores = tf.reduce_sum(avg_embedding * movie_embeddings, axis=1).numpy()
        top_indices = np.argsort(scores)[::-1]

        # Top N recommendations excluding already-liked
        recommended = []
        for idx in top_indices:
            mid = reverse_movie_map.get(idx)
            if mid not in user_ratings and mid in movie_titles:
                recommended.append((movie_titles[mid], scores[idx]))
            if len(recommended) >= 10:
                break

        # Display recommendations
        st.subheader("🍿 Top 10 Recommendations")
        for title, score in recommended:
            st.write(f"**{title}** — Score: `{score:.3f}`")