app.py

import requests
import pandas as pd
import spacy
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
from textblob import TextBlob
import streamlit as st

# 🔹 Replace this with your TMDB API Key
API_KEY = "bbb69cf69be036e363d9ab8996f7f4ee"
BASE_URL = "https://api.themoviedb.org/3"

# 🔹 TMDB Image Base URL for posters
IMAGE_BASE_URL = "https://image.tmdb.org/t/p/w500"

# Load Spacy English NLP model
nlp = spacy.load("en_core_web_sm")


# ===========================
#  STEP 1: FETCH MOVIE DATA
# ===========================
def fetch_movies(num_pages=2):
    """Fetch popular movies from TMDB API."""
    all_movies = []

    for page in range(1, num_pages + 1):
        url = f"{BASE_URL}/discover/movie?api_key={API_KEY}&language=en-US&sort_by=popularity.desc&page={page}"
        response = requests.get(url)
        data = response.json()

        if "results" in data:
            for movie in data["results"]:
                all_movies.append({
                    "id": movie["id"],
                    "title": movie["title"],
                    "overview": movie["overview"],
                    "vote_average": movie["vote_average"],
                    "release_date": movie["release_date"]
                })

    return pd.DataFrame(all_movies)


# ===========================
#  STEP 2: FETCH ADDITIONAL DETAILS (GENRES, CAST, DIRECTOR)
# ===========================
def fetch_genres():
    """Retrieve genre names from TMDB API and return a dictionary mapping genre IDs to names."""
    url = f"{BASE_URL}/genre/movie/list?api_key={API_KEY}&language=en-US"
    response = requests.get(url)
    data = response.json()

    return {genre["id"]: genre["name"] for genre in data["genres"]}


def fetch_movie_details(movie_id):
    """Fetch top 3 cast members and director for a given movie."""
    url = f"{BASE_URL}/movie/{movie_id}/credits?api_key={API_KEY}"
    response = requests.get(url)
    data = requests.get(url).json()

    # Get top 3 cast members
    cast = ", ".join([member["name"] for member in data.get("cast", [])[:3]])

    # Get director
    director = next((crew["name"] for crew in data.get("crew", []) if crew["job"] == "Director"), "Unknown")

    return cast, director


# ===========================
#  STEP 3: ENRICH MOVIE DATA WITH GENRES, CAST, DIRECTOR
# ===========================
def enhance_movie_data(movies_df):
    """Add genres, top cast, and director information to the movie dataset."""
    genre_dict = fetch_genres()

    movies_df["cast"], movies_df["director"] = zip(*movies_df["id"].apply(fetch_movie_details))
    return movies_df


# ===========================
#  STEP 4: FEATURE ENGINEERING (KEYWORDS & SENTIMENT)
# ===========================
def extract_keywords(text, num_keywords=5):
    """Extract top keywords from text using TF-IDF."""
    vectorizer = TfidfVectorizer(stop_words="english", max_features=50)
    tfidf_matrix = vectorizer.fit_transform([text])
    feature_names = vectorizer.get_feature_names_out()

    sorted_indices = tfidf_matrix.toarray().argsort()[0][-num_keywords:]
    return ", ".join(feature_names[i] for i in sorted_indices)


def get_sentiment(text):
    """Analyze sentiment (-1 to 1) from movie description using TextBlob."""
    return TextBlob(text).sentiment.polarity


# ===========================
#  STEP 5: BUILD RECOMMENDER SYSTEM (CONTENT-BASED FILTERING)
# ===========================
def recommend_movies(movie_title, num_recommendations=5):
    """Recommend similar movies based on content similarity."""
    if movie_title not in movies_df["title"].values:
        return "Movie not found in dataset!"

    # Get index of the selected movie
    movie_index = movies_df[movies_df["title"] == movie_title].index[0]

    # Compute similarity scores and sort them
    similarity_scores = list(enumerate(cosine_sim[movie_index]))
    similarity_scores = sorted(similarity_scores, key=lambda x: x[1], reverse=True)[1:num_recommendations + 1]

    # Get recommended movie titles
    return [movies_df.iloc[i[0]]["title"] for i in similarity_scores]


# ===========================
#  STEP 6: FETCH MOVIE POSTER
# ===========================
def get_movie_poster(movie_title):
    """Fetch movie poster from TMDB API."""
    movie = movies_df[movies_df["title"] == movie_title]
    if not movie.empty:
        movie_id = movie.iloc[0]["id"]
        url = f"https://api.themoviedb.org/3/movie/{movie_id}?api_key={API_KEY}"
        response = requests.get(url).json()
        return IMAGE_BASE_URL + response.get("poster_path", "")
    return None


# ===========================
#  STEP 7: LOAD & PROCESS MOVIE DATA
# ===========================
movies_df = fetch_movies(num_pages=2)  # Fetch movie data
movies_df = enhance_movie_data(movies_df)  # Add cast, director info

# Apply feature extraction
movies_df["keywords"] = movies_df["overview"].apply(lambda x: extract_keywords(str(x)))
movies_df["sentiment"] = movies_df["overview"].apply(lambda x: get_sentiment(str(x)))

# Combine relevant text features for recommendation
movies_df["combined_features"] = (
        movies_df["overview"].fillna("") + " " +
        movies_df["keywords"].fillna("")
)

# Convert text into numerical vectors using TF-IDF
tfidf_vectorizer = TfidfVectorizer(stop_words="english")
tfidf_matrix = tfidf_vectorizer.fit_transform(movies_df["combined_features"])

# Compute similarity scores between movies
cosine_sim = cosine_similarity(tfidf_matrix, tfidf_matrix)

# ===========================
#  STEP 8: STREAMLIT APP UI
# ===========================
st.title("🎬 Movie Recommendation System")

# Dropdown to select a movie
selected_movie = st.selectbox("Select a Movie", movies_df["title"].values)

# Recommend button
if st.button("Recommend"):
    recommendations = recommend_movies(selected_movie)

    if isinstance(recommendations, list):
        st.subheader(f"Movies similar to {selected_movie}:")

        # Display recommended movies in a horizontal layout
        cols = st.columns(len(recommendations))

        for i, movie in enumerate(recommendations):
            poster_url = get_movie_poster(movie)
            with cols[i]:
                if poster_url:
                    st.image(poster_url, width=150)
                st.write(f"**{movie}**")
    else:
        st.error("No recommendations found.")