src/streamlit_app.py

import streamlit as st
import joblib
import pandas as pd
import os # Operating system module for path manipulation

# Define the directory where the models are stored
# File Names
MODEL_FILENAME = 'src/music_model.joblib'
SCALER_FILENAME = 'src/music_scaler.joblib'

# --- Load Saved Models ---
try:
    # Use os.path.join to create the correct file paths
    loaded_model = joblib.load(MODEL_FILENAME)
    loaded_scaler = joblib.load(SCALER_FILENAME)
    
except FileNotFoundError:
    st.error(f"Model or Scaler files not found! Please ensure 'music_model.joblib' and 'music_scaler.joblib' are located in the '{MODEL_DIR}' directory.")
    st.stop()
except Exception as e:
    st.error(f"An error occurred while loading the models: {e}")
    st.stop()


# --- Streamlit Interface ---
st.title("🎧 Music Clustering Prediction App")
st.markdown("Enter the song's audio features to see which music segment it belongs to!")

# Get User Input (Ensure the feature names match the training data order)
st.subheader("Enter Song Audio Features:")

bpm = st.slider("Beats Per Minute (BPM)", min_value=50, max_value=200, value=120)
loudness = st.slider("Loudness (dB)", min_value=-20.0, max_value=0.0, value=-5.0, format="%.2f")
liveness = st.slider("Liveness", min_value=0.0, max_value=1.0, value=0.5, format="%.2f")
valence = st.slider("Valence (Positivity/Cheerfulness)", min_value=0.0, max_value=1.0, value=0.6, format="%.2f")
acousticness = st.slider("Acousticness", min_value=0.0, max_value=1.0, value=0.2, format="%.2f")
speechiness = st.slider("Speechiness", min_value=0.0, max_value=1.0, value=0.1, format="%.2f")


# Define feature names in the exact order used during training/scaling
FEATURE_NAMES = ['Beats Per Minute (BPM)', 'Loudness (dB)', 'Liveness', 
                 'Valence', 'Acousticness', 'Speechiness']


if st.button('Perform Cluster Prediction'):
    # 1. Convert inputs into a DataFrame
    new_data = pd.DataFrame([[bpm, loudness, liveness, valence, acousticness, speechiness]], 
                            columns=FEATURE_NAMES)
    
    # 2. Transform the new data using the loaded scaler
    scaled_new_data = loaded_scaler.transform(new_data)
    
    # 3. Make the prediction using the model
    prediction = loaded_model.predict(scaled_new_data)
    
    # 4. Display the result
    st.success(f"Based on its audio characteristics, this song belongs to **Cluster {prediction[0]}**!")