import streamlit as st
import pandas as pd
from sklearn.preprocessing import LabelEncoder
import tensorflow as tf
from tensorflow import keras
from keras.metrics import Precision, Recall
import numpy as np

# ✅ Set page config — must be the very first Streamlit call
st.set_page_config(page_title="💖 Heartify", layout="centered", page_icon="💖")

# ✅ CSS Overrides to remove blur/spinner effects
st.markdown(
    """
    <style>
    [data-testid="stAppViewContainer"] * {
        filter: none !important;
        -webkit-filter: none !important;
    }
    div[data-testid="stSpinner"]::before {
        background: none !important;
    }
    [data-baseweb="checkbox"] * {
        filter: none !important;
        -webkit-filter: none !important;
        opacity: 1 !important;
    }
    [data-baseweb="input"] * {
        filter: none !important;
        -webkit-filter: none !important;
        opacity: 1 !important;
    }
    .block-container:has(div[aria-busy="true"]) {
        pointer-events: auto !important;
    }
    </style>
    """,
    unsafe_allow_html=True
)

# ✅ Load dataset
df = pd.read_csv("data.csv")

# ✅ Drop unnecessary columns
for col in ["Timestamp", "Email Address"]:
    if col in df.columns:
        df.drop(columns=col, inplace=True)

# ✅ Label Encoding
encoders = {}
for col in df.columns:
    le = LabelEncoder()
    df[col] = le.fit_transform(df[col])
    encoders[col] = le

# ✅ Features and labels
x = df.iloc[:, :-1]
y = df.iloc[:, -1]

# ✅ Build ANN model
model = keras.Sequential([
    keras.layers.Input(shape=(x.shape[1],)),
    keras.layers.Dense(8, kernel_initializer=keras.initializers.GlorotNormal(seed=42)),
    keras.layers.PReLU(),
    keras.layers.Dense(3, kernel_initializer=keras.initializers.GlorotNormal(seed=42),
                       kernel_regularizer=keras.regularizers.L2()),
    keras.layers.PReLU(),
    keras.layers.BatchNormalization(),
    keras.layers.Dropout(0.2),
    keras.layers.Dense(1, activation="sigmoid", kernel_initializer=keras.initializers.HeNormal(seed=42))
])

model.compile(loss="binary_crossentropy", optimizer="adam", metrics=["accuracy", Precision(), Recall()])

# ✅ App Header
st.title("💖 Heartify")
st.markdown("#### *Let your heart meet its algorithm*")
st.caption("💬 హృదయాలను అనుసంధానించే ఏకైక AI - Heartify")
st.divider()

# ✅ Train model on button click
if st.button("✨ Prepare the Heartify AI"):
    with st.spinner("🔄 Training the model... please wait."):
        model.fit(x, y, validation_split=0.2, batch_size=8, epochs=30, verbose=0)
    st.success("✅ Heartify is ready to make predictions!")

# ✅ User Input
st.subheader("📋 Tell Us About Yourself")
user_inputs = {}
for col in x.columns:
    options = encoders[col].classes_
    selected = st.selectbox(f"🔹 {col.replace('_', ' ').title()}", options)
    user_inputs[col] = encoders[col].transform([selected])[0]

# ✅ Predict Compatibility
if st.button("💌 Check Compatibility"):
    input_df = pd.DataFrame([user_inputs.values()], columns=user_inputs.keys())
    prediction = model.predict(input_df)[0][0]

    st.markdown("---")
    if prediction >= 0.5:
        st.success("💘 It's a Match! Your vibes align perfectly 💞")
    else:
        st.warning("🕊️ Not quite yet — love takes time. Keep believing.")