import streamlit as st
import joblib
import pandas as pd
from sklearn.metrics.pairwise import cosine_similarity

tfidf = joblib.load("src/tfidf_vectorizer.pkl")
df = pd.read_csv("src/metadata.csv")

tfidf_matrix = tfidf.transform(df["combined_features"])
cosine_sim = cosine_similarity(tfidf_matrix, tfidf_matrix)

indices = pd.Series(df.index, index=df["track_name"]).drop_duplicates()

def recommend(song_name, n=5):
    idx = indices[song_name]
    idx = idx.iloc[0] if hasattr(idx, "iloc") else idx

    sim_scores = list(enumerate(cosine_sim[idx]))
    sim_scores = sorted(sim_scores, key=lambda x: x[1], reverse=True)[1:n+1]

    song_indices = [i[0] for i in sim_scores]
    return df.loc[song_indices, ["track_name", "track_artist", "playlist_genre"]]

st.title("🎵 Spotify Recommendation System")

song = st.selectbox("Bir şarkı seçin:", df["track_name"].unique())

if st.button("Öner"):
    recommendations = recommend(song)
    st.dataframe(recommendations)