import streamlit as st
import os
import pandas as pd
import requests
from sentence_transformers import SentenceTransformer
from sklearn.metrics.pairwise import cosine_similarity


st.set_page_config(page_title="AI Movie Recommender", layout="wide")

st.title("🎬 AI Movie Recommendation System")


# ---------------------------
# Load movie dataset
# ---------------------------

data = {
    "title": [
        "The Dark Knight",
        "Batman Begins",
        "Interstellar",
        "Inception",
        "The Matrix",
        "John Wick",
        "The Notebook",
        "Titanic",
        "Avengers: Endgame",
        "Iron Man"
    ],

    "description": [
        "Batman fights the Joker to save Gotham City",
        "Bruce Wayne becomes Batman to fight crime",
        "Astronauts travel through a wormhole to save humanity",
        "A thief enters dreams to steal secrets",
        "A hacker discovers reality is a simulation",
        "An assassin seeks revenge for his dog",
        "A romantic love story between two young people",
        "A tragic love story on the Titanic ship",
        "Superheroes unite to defeat Thanos",
        "A billionaire builds a high-tech armored suit"
    ]
}

movies = pd.DataFrame(data)


# ---------------------------
# Load pretrained model
# ---------------------------

model = SentenceTransformer("all-MiniLM-L6-v2")


# ---------------------------
# Generate embeddings
# ---------------------------

embeddings = model.encode(movies["description"])


# ---------------------------
# Similarity matrix
# ---------------------------

similarity_matrix = cosine_similarity(embeddings)


# ---------------------------
# Fetch movie poster
# ---------------------------


def fetch_poster(movie_title):

    api_key = os.getenv("My_TMDB_KEY")

    if not api_key:
        return None

    url = f"https://api.themoviedb.org/3/search/movie?api_key={api_key}&query={movie_title}"

    try:
        response = requests.get(url, timeout=5)

        if response.status_code != 200:
            return None

        data = response.json()

        # Safe validation
        results = data.get("results", [])

        if len(results) == 0:
            return None

        poster_path = results[0].get("poster_path")

        if not poster_path:
            return None

        return f"https://image.tmdb.org/t/p/w500/{poster_path}"

    except Exception as e:
        return None

# ---------------------------
# Recommendation function
# ---------------------------

def recommend(movie_title):

    idx = movies[movies["title"] == movie_title].index[0]

    similarity_scores = list(enumerate(similarity_matrix[idx]))

    similarity_scores = sorted(similarity_scores, key=lambda x: x[1], reverse=True)

    similarity_scores = similarity_scores[1:6]

    movie_indices = [i[0] for i in similarity_scores]

    return movies.iloc[movie_indices]["title"].values


# ---------------------------
# Search UI
# ---------------------------

selected_movie = st.selectbox(
    "Search a movie",
    movies["title"].values
)


if st.button("Recommend"):

    recommended_movies = recommend(selected_movie)

    cols = st.columns(5)

    for i, movie in enumerate(recommended_movies):

        poster = fetch_poster(movie)

        with cols[i]:

            st.text(movie)

            if poster:
                st.image(poster)