app.py

import streamlit as st
import requests
from transformers import pipeline

# Enter your TMDb API key here
TMDB_API_KEY = '2cc77b08a475b510f54a0dcca2c5c0c7'
IMAGE_BASE_URL = "https://image.tmdb.org/t/p/w200"  # TMDb base URL for poster images (width 200)

@st.cache_resource
def load_emotion_analyzer():
    return pipeline("text-classification", model="j-hartmann/emotion-english-distilroberta-base")

emotion_analyzer = load_emotion_analyzer()

def analyze_emotion(description):
    """Analyze emotion of a movie's description."""
    emotions = emotion_analyzer(description)
    return emotions[0]['label']  # Return the top emotion

def get_movie_genres():
    """Fetches movie genres from TMDb API and returns a dictionary of genres."""
    url = f'https://api.themoviedb.org/3/genre/movie/list?api_key={TMDB_API_KEY}&language=en-US'
    response = requests.get(url)
    if response.status_code == 200:
        genres = response.json().get('genres', [])
        return {genre['name']: genre['id'] for genre in genres}
    else:
        return None  # Return None if the request failed

def get_top_movies_by_genre(genre_id):
    """Fetches the top 10 highest-rated movies for a given genre ID from TMDb, sorted by title length."""
    url = f'https://api.themoviedb.org/3/discover/movie?api_key={TMDB_API_KEY}&language=en-US&sort_by=vote_average.desc&vote_count.gte=1000&with_genres={genre_id}'
    response = requests.get(url)
    if response.status_code == 200:
        movies = response.json().get('results', [])[:10]  # Get the top 10 results
        # Sort movies by title length, shortest to longest
        return sorted([(movie['title'], movie['vote_average'], movie['id'], movie['poster_path']) for movie in movies], key=lambda x: len(x[0]))
    else:
        return None

def get_similar_movies(movie_id, genre_id):
    """Fetches similar movies for a given movie ID from TMDb and filters by the selected genre."""
    url = f'https://api.themoviedb.org/3/movie/{movie_id}/similar?api_key={TMDB_API_KEY}&language=en-US'
    response = requests.get(url)
    if response.status_code == 200:
        movies = response.json().get('results', [])[:15]
        # Include the movie description (overview) in the returned values
        return [
            (movie['title'], movie['vote_average'], movie['poster_path'], movie['id'], movie['overview'])
            for movie in movies if genre_id in movie['genre_ids'] and 'overview' in movie
        ]
    else:
        return None

def get_movie_details(movie_id):
    """Fetches detailed information for a given movie ID from TMDb, including description, release date, director, and cast."""
    url = f'https://api.themoviedb.org/3/movie/{movie_id}?api_key={TMDB_API_KEY}&language=en-US&append_to_response=credits'
    response = requests.get(url)
    if response.status_code == 200:
        movie = response.json()
        directors = [member['name'] for member in movie['credits']['crew'] if member['job'] == 'Director']
        cast = [member['name'] for member in movie['credits']['cast'][:5]]  # Get top 5 cast members
        return {
            "title": movie['title'],
            "rating": movie['vote_average'],
            "description": movie['overview'],
            "poster_path": movie['poster_path'],
            "release_date": movie['release_date'],
            "director": ", ".join(directors),
            "cast": ", ".join(cast)
        }
    else:
        return None

# Initialize session state to store selected movies and recommendations
if 'selected_movies' not in st.session_state:
    st.session_state['selected_movies'] = set()
if 'recommendations' not in st.session_state:
    st.session_state['recommendations'] = []
if 'last_genre' not in st.session_state:
    st.session_state['last_genre'] = None

# Streamlit app title
st.title("Movie Recommender")

# CSS for alignment and button styling
st.markdown("""
    <style>
    .movie-container {
        display: flex;
        flex-direction: column;
        align-items: center;
        justify-content: flex-start;
    }
    .movie-caption {
        min-height: 50px;  /* Set a consistent height for caption sections */
        display: flex;
        align-items: center;
        justify-content: center;
        text-align: center;
    }
    .movie-caption button {
        width: 100%;
    }
    </style>
""", unsafe_allow_html=True)

# Fetch genres from TMDb and display in a dropdown
genres = get_movie_genres()
if genres is None:
    st.error("Failed to load movie genres. Please check your API key or connection.")
else:
    genre_name = st.selectbox("Select a movie genre:", [""] + list(genres.keys()))  # Include an empty option

    # Check if genre has changed, and reset selections if so
    if genre_name != st.session_state['last_genre']:
        st.session_state['selected_movies'].clear()
        st.session_state['recommendations'].clear()
        st.session_state['last_genre'] = genre_name

    # Only fetch and display movies if a genre is selected
    if genre_name:
        st.write("Please select at least one movie to generate recommendations.")
        genre_id = genres[genre_name]  # Get the genre ID based on the selected genre name
        top_movies = get_top_movies_by_genre(genre_id)  # Fetch top 10 movies in this genre, sorted by title length

        # Check if movies were successfully loaded
        if top_movies is None:
            st.error("Failed to load movies. Please check your API key or connection.")
        else:
            # Display the top 10 movies as clickable titles in a 5x2 grid layout with aligned rows
            st.write(f"Top 10 movies in the {genre_name} genre:")
            columns = st.columns(5)  # Create 5 columns for each row

            # First row of 5 movies
            for index, (title, _, movie_id, poster_path) in enumerate(top_movies[:5]):
                col = columns[index]
                with col:
                    st.markdown("<div class='movie-container'>", unsafe_allow_html=True)

                    # Display image
                    st.image(f"{IMAGE_BASE_URL}{poster_path}", use_container_width=True)

                    # Display clickable title as a button to toggle selection, with consistent height
                    if st.button(f"{title} {'✅' if movie_id in st.session_state['selected_movies'] else ''}", key=f"title_{movie_id}"):
                        if movie_id in st.session_state['selected_movies']:
                            st.session_state['selected_movies'].remove(movie_id)
                        else:
                            st.session_state['selected_movies'].add(movie_id)

                    st.markdown("</div>", unsafe_allow_html=True)

            # Second row of 5 movies
            for index, (title, _, movie_id, poster_path) in enumerate(top_movies[5:]):
                col = columns[index]
                with col:
                    st.markdown("<div class='movie-container'>", unsafe_allow_html=True)

                    # Display image
                    st.image(f"{IMAGE_BASE_URL}{poster_path}", use_container_width=True)

                    # Display clickable title as a button to toggle selection, with consistent height
                    if st.button(f"{title} {'✅' if movie_id in st.session_state['selected_movies'] else ''}", key=f"title_2_{movie_id}"):
                        if movie_id in st.session_state['selected_movies']:
                            st.session_state['selected_movies'].remove(movie_id)
                        else:
                            st.session_state['selected_movies'].add(movie_id)

                    st.markdown("</div>", unsafe_allow_html=True)

    # Button to generate recommendations based on selected movies
    if st.button("Show Recommendations"):
        if st.session_state['selected_movies']:
            # Step 1: Collect emotions of user-selected movies
            with st.spinner("Analyzing emotions of selected movies..."):
                selected_emotions = []
                for movie_id in st.session_state['selected_movies']:
                    movie_details = get_movie_details(movie_id)
                    if movie_details:
                        emotion = analyze_emotion(movie_details["description"])
                        selected_emotions.append(emotion)

                # Step 2: Collect similar movies for all selected movies
                all_similar_movies = []
                for movie_id in st.session_state['selected_movies']:
                    similar_movies = get_similar_movies(movie_id, genre_id)
                    if similar_movies:
                        all_similar_movies.extend(similar_movies)

                # Step 3: Filter similar movies by matching emotions and remove duplicates
                recommendations = []
                seen_movie_ids = set(st.session_state['selected_movies'])  # Start with user-selected movies to exclude them

            with st.spinner("Filtering movies by description emotion..."):
                for title, rating, poster_path, movie_id, description in all_similar_movies:
                    if movie_id not in seen_movie_ids:  # Exclude already-seen movies
                        movie_emotion = analyze_emotion(description)
                        if movie_emotion in selected_emotions:
                            seen_movie_ids.add(movie_id)  # Mark this movie as seen to prevent duplicates
                            recommendations.append((title, rating, poster_path, movie_id, movie_emotion))

            # Store unique recommendations in session state
            st.session_state['recommendations'] = recommendations

            # Step 4: Display final recommendations
            if st.session_state['recommendations']:
                st.write("Based on your selections, you might also enjoy these top movies:")
                for title, rating, poster_path, movie_id, emotion in sorted(st.session_state['recommendations'], key=lambda x: x[1], reverse=True)[:5]:
                    movie_details = get_movie_details(movie_id)
                    if movie_details:
                        col1, col2 = st.columns([1, 3])
                        with col1:
                            st.image(f"{IMAGE_BASE_URL}{movie_details['poster_path']}", use_container_width=True)
                        with col2:
                            st.markdown(f"**Title:** {movie_details['title']}")
                            st.markdown(f"**TMDb Rating:** {movie_details['rating']}")
                            st.markdown(f"**Description:** {movie_details['description']}")

                            # Additional Information button with overlay using expander
                            with st.expander("Additional Information"):
                                st.markdown(f"**Release Date:** {movie_details['release_date']}")
                                st.markdown(f"**Director:** {movie_details['director']}")
                                st.markdown(f"**Cast:** {movie_details['cast']}")
            else:
                st.warning("No movies matched your selected emotions.")