src/streamlit_app.py

import numpy as np
import pandas as pd
import streamlit as st
import joblib
from pathlib import Path

# -------------------------------
# Page config
# -------------------------------
st.set_page_config(
    page_title='Rainfall Probability Predictor (LogReg)',
    page_icon='🌧️',
    layout='centered'
)

st.title('🌧️ Rainfall Probability Predictor')
st.write('Predicts the probability of rainfall (0–1) using a Logistic Regression model trained on weather features.')

BASE_DIR = Path(__file__).resolve().parent
MODEL_PATH = BASE_DIR / 'lr_final_model.pkl'
FEATURE_NAMES_PATH = BASE_DIR / 'feature_names.pkl'


@st.cache_resource
def load_artifacts():
    if not MODEL_PATH.exists():
        raise FileNotFoundError(
            f'Model not found: {MODEL_PATH.name}. Put it in the repo root (same folder as app.py).'
        )
    if not FEATURE_NAMES_PATH.exists():
        raise FileNotFoundError(
            f'Feature names not found: {FEATURE_NAMES_PATH.name}. Put it in the repo root (same folder as app.py).'
        )

    model = joblib.load(MODEL_PATH)
    feature_names = joblib.load(FEATURE_NAMES_PATH)

    if not isinstance(feature_names, list) or len(feature_names) == 0:
        raise ValueError('feature_names.pkl must contain a non-empty list of column names.')

    return model, feature_names


def add_features(df: pd.DataFrame) -> pd.DataFrame:
    df = df.copy()

    # Temperature variability
    df['temp_range'] = df['maxtemp'] - df['mintemp']
    # Air saturation level
    df['humidity_gap'] = df['humidity'] - df['dewpoint']
    # Sunshine vs clouds
    df['sunshine_ratio'] = df['sunshine'] / (df['cloud'] + 1)
    # Wind intensity (simple interaction)
    df['wind_energy'] = df['windspeed'] * df['winddirection']
    # Seasonal pattern
    df['sin_day'] = np.sin(2 * np.pi * df['day'] / 365)
    df['cos_day'] = np.cos(2 * np.pi * df['day'] / 365)

    return df


model, feature_names = load_artifacts()

st.subheader('Input features')

# Note: ranges are generic. If you want, you can set them based on df1.describe().
col1, col2 = st.columns(2)

with col1:
    day = st.number_input('day (1–365)', min_value=1, max_value=365, value=100, step=1)
    pressure = st.number_input('pressure', value=1013.0, step=0.1)
    maxtemp = st.number_input('maxtemp', value=20.0, step=0.1)
    temperature = st.number_input('temperature', value=15.0, step=0.1)
    mintemp = st.number_input('mintemp', value=10.0, step=0.1)

with col2:
    dewpoint = st.number_input('dewpoint', value=8.0, step=0.1)
    humidity = st.number_input('humidity', value=70.0, step=0.1)
    cloud = st.number_input('cloud', value=50.0, step=1.0)
    sunshine = st.number_input('sunshine', value=5.0, step=0.1)
    windspeed = st.number_input('windspeed', value=10.0, step=0.1)
    winddirection = st.number_input('winddirection', value=180.0, step=1.0)

# Build one-row dataframe with the ORIGINAL base features
input_df = pd.DataFrame([{
    'day': float(day),
    'pressure': float(pressure),
    'maxtemp': float(maxtemp),
    'temparature': float(temperature),
    'mintemp': float(mintemp),
    'dewpoint': float(dewpoint),
    'humidity': float(humidity),
    'cloud': float(cloud),
    'sunshine': float(sunshine),
    'windspeed': float(windspeed),
    'winddirection': float(winddirection)
}])


# Add engineered features (must match training)
input_df = add_features(input_df)

# Ensure correct feature order and missing columns safety
missing_cols = [c for c in feature_names if c not in input_df.columns]
extra_cols = [c for c in input_df.columns if c not in feature_names]

if missing_cols:
    st.error(f'Missing required feature columns: {missing_cols}')
    st.stop()

# Keep only the expected columns in the correct order
X = input_df[feature_names].copy()

st.divider()

if st.button('Predict rainfall probability'):
    try:
        proba = float(model.predict_proba(X)[:, 1][0])
        st.metric('Rainfall probability', f'{proba:.3f}', delta=None)
        st.progress(min(max(proba, 0.0), 1.0))

        if proba >= 0.7:
            st.success('High chance of rainfall.')
        elif proba >= 0.4:
            st.warning('Medium chance of rainfall.')
        else:
            st.info('Low chance of rainfall.')

        with st.expander('Show model input (debug)'):
            st.write('Used feature columns (ordered):')
            st.write(feature_names)
            st.dataframe(X)

        if extra_cols:
            st.caption(f'Note: These columns were ignored (not in feature_names): {extra_cols}')

    except Exception as e:
        st.error(f'Prediction failed: {e}')