app.py

import streamlit as st
import pandas as pd
import joblib
import tempfile
import os
from train_models import run_automl
import plotly.express as px

st.set_page_config(page_title="🔮 AutoML Explorer", layout="wide")
st.title("🔮 AutoML Explorer - Predict Anything From Any CSV")
st.markdown("""
Welcome to **AutoML Explorer** — your no-code ML assistant:
- 📂 Upload any CSV
- 🧠 Let the app clean, preprocess & analyze your data
- 🤖 Tries many ML models including XGBoost & LightGBM
- 🎯 Shows the best with full explanation & visuals
- 🔮 Make live predictions & download result CSV
""")

file = st.file_uploader("📁 Upload your CSV file", type=["csv"])

if file:
    df = pd.read_csv(file)
    st.subheader("📊 Dataset Preview")
    st.dataframe(df.head())

    st.subheader("🎯 Choose Your Target Column")
    target = st.selectbox("Select what you want to predict:", df.columns)

    if st.button("🚀 Run AutoML"):
        with st.spinner("🤖 Running smart ML analysis..."):
            result = run_automl(df, target)

        if "error" in result:
            st.error(result["error"])
        else:
            st.success("✨ AutoML Complete! Here's your custom ML report:")

            st.markdown("### 🔍 What We Did With Your Data")
            st.json(result["preprocessing"])

            st.markdown("### 🤖 Model Scores")
            model_df = pd.DataFrame(result["model_scores"], index=["Score (%)"]).T
            st.dataframe(model_df)

            # Bar Chart
            st.markdown("### 📊 Visual Model Comparison")
            chart = px.bar(model_df, x=model_df.index, y="Score (%)", title="Model Performance Comparison", color_discrete_sequence=["#636EFA"])
            st.plotly_chart(chart, use_container_width=True)

            st.markdown("### ✅ Best Model Chosen")
            st.markdown(f"🏆 **{result['best_model']}** with **{result['best_accuracy']:.2f}% accuracy/performance**")

            if result["type"] == "classification":
                st.markdown("### 📈 Confusion Matrix")
                st.dataframe(pd.DataFrame(result["confusion_matrix"]))

                st.markdown("### 🧾 Classification Report")
                st.json(result["classification_report"])
            else:
                st.markdown("### 📏 Regression Metrics")
                st.json(result["regression_report"])

            # Save model & create download button
            with tempfile.NamedTemporaryFile(delete=False, suffix=".pkl") as tmp_file:
                joblib.dump(result["model_object"], tmp_file.name)
                st.download_button("📥 Download Best Model (.pkl)", data=open(tmp_file.name, "rb"), file_name="best_model.pkl")

            # Predictions CSV
            st.markdown("### 📄 Prediction Results (Test Set)")
            st.dataframe(result["predictions"])
            csv = result["predictions"].to_csv(index=False).encode('utf-8')
            st.download_button("📥 Download Predictions CSV", csv, "predictions.csv", "text/csv")

            # Live prediction input
            st.markdown("### 🔮 Live Prediction Input")
            input_data = {}
            for col in df.drop(columns=[target]).columns:
                dtype = df[col].dtype
                if dtype == 'object':
                    input_data[col] = st.selectbox(f"{col}", options=sorted(df[col].dropna().unique()))
                else:
                    input_data[col] = st.number_input(f"{col}", value=float(df[col].mean()))

            if st.button("🎯 Predict with Best Model"):
                user_df = pd.DataFrame([input_data])
                from utils import preprocess_data
                try:
                    X_temp, _, _, _ = preprocess_data(pd.concat([df, user_df], ignore_index=True), target)
                    user_input = X_temp.iloc[-1:]
                    prediction = result["model_object"].predict(user_input)[0]
                    st.success(f"🧠 Predicted Output: {prediction}")
                except Exception as e:
                    st.error(f"Prediction failed: {e}")