src/streamlit_app.py

import streamlit as st
import pandas as pd
import joblib
import matplotlib.pyplot as plt
import os

# --------------------------------------
# 🎨 PAGE CONFIG
# --------------------------------------
st.set_page_config(page_title="Spotify Clustering", layout="wide")

# --------------------------------------
# 🔧 AUTO PATH (LOCAL + CLOUD FIX)
# --------------------------------------
def get_path(filename):
    base_dir = os.path.dirname(os.path.abspath(__file__))

    local_path = os.path.join(base_dir, filename)
    parent_path = os.path.join(base_dir, "..", filename)

    if os.path.exists(local_path):
        return local_path
    elif os.path.exists(parent_path):
        return parent_path
    else:
        st.error(f"{filename} bulunamadı!")
        st.stop()

DATA_PATH = get_path("spotify_clustered.csv")
MODEL_PATH = get_path("kmeans_music_model.pkl")
SCALER_PATH = get_path("scaler_music.pkl")

# --------------------------------------
# 📂 LOAD DATA (CACHE)
# --------------------------------------
@st.cache_data
def load_data():
    df = pd.read_csv(DATA_PATH)

    unnecessary_cols = ['Unnamed: 0', 'track_id', 'album_name', 'explicit']
    df_display = df.drop(columns=[c for c in unnecessary_cols if c in df.columns])

    return df, df_display

@st.cache_resource
def load_model():
    model = joblib.load(MODEL_PATH)
    scaler = joblib.load(SCALER_PATH)
    return model, scaler

df, df_display = load_data()
model, scaler = load_model()

# --------------------------------------
# 🧠 SESSION STATE (NO FLICKER)
# --------------------------------------
if "prediction" not in st.session_state:
    st.session_state.prediction = None
    st.session_state.samples = None

# --------------------------------------
# 🏷️ TITLE
# --------------------------------------
st.title("🎵 Spotify Music Clustering / Spotify Müzik Kümeleme")
st.markdown("---")

# --------------------------------------
# 📄 DATA PREVIEW
# --------------------------------------
st.subheader("📄 Dataset Preview / Veri Önizleme")
st.dataframe(df_display.head(10), use_container_width=True)

# --------------------------------------
# 📊 VISUALS
# --------------------------------------
col1, col2 = st.columns(2)

with col1:
    st.subheader("📊 Cluster Distribution / Küme Dağılımı")
    st.bar_chart(df['cluster'].value_counts(), use_container_width=True)

with col2:
    st.subheader("🎯 Feature Analysis / Özellik Analizi")

    fig, ax = plt.subplots(figsize=(6, 4))
    scatter = ax.scatter(
        df['danceability'],
        df['energy'],
        c=df['cluster'],
        cmap='viridis',
        alpha=0.6
    )

    ax.set_xlabel("Danceability / Dans Edilebilirlik")
    ax.set_ylabel("Energy / Enerji")

    plt.colorbar(scatter, ax=ax)

    st.pyplot(fig, clear_figure=True)
    plt.close(fig)

# --------------------------------------
# 🤖 PREDICTION
# --------------------------------------
st.divider()
st.subheader("🤖 Predict New Song Cluster / Yeni Şarkı Tahmini")

c1, c2, c3 = st.columns(3)

with c1:
    pop = st.slider("Popularity / Popülerlik", 0, 100, 50)
    dur = st.slider("Duration (ms) / Süre", 0, 600000, 200000)
    dance = st.slider("Danceability / Dans Edilebilirlik", 0.0, 1.0, 0.5)

with c2:
    energy = st.slider("Energy / Enerji", 0.0, 1.0, 0.5)
    loud = st.slider("Loudness / Ses Yüksekliği", -60.0, 0.0, -10.0)
    tempo = st.slider("Tempo / Tempo", 0.0, 250.0, 120.0)

with c3:
    speech = st.slider("Speechiness / Konuşma Oranı", 0.0, 1.0, 0.1)

if st.button("Predict Cluster / Kümeyi Tahmin Et ✨"):
    new_data = [[pop, dur, dance, energy, loud, tempo, speech]]

    try:
        new_data_scaled = scaler.transform(new_data)
        res = model.predict(new_data_scaled)[0]

        st.session_state.prediction = res
        st.session_state.samples = df[df['cluster'] == res][['track_name', 'artists']].head(5)

    except Exception as e:
        st.error(f"Error / Hata: {e}")

# SONUÇ (STABLE)
if st.session_state.prediction is not None:
    st.success(f"### Predicted Cluster / Tahmin Edilen Küme: {st.session_state.prediction}")
    st.write("Similar Songs / Benzer Şarkılar:")
    st.table(st.session_state.samples)

# --------------------------------------
# 🔍 CLUSTER ANALYSIS
# --------------------------------------
st.divider()
st.subheader("🔍 Cluster Characteristics / Küme Özellikleri")

numeric_only = df.select_dtypes(include=['float64', 'int64'])

if 'cluster' in df.columns:
    means = numeric_only.groupby(df['cluster']).mean()
    st.dataframe(means, use_container_width=True)