app.py

import streamlit as st
import pandas as pd
import os
from ml_engine import PredictIQEngine
import base64
from dotenv import load_dotenv

# Load environment variables
load_dotenv()

# Set Page Config
st.set_page_config(
    page_title="PredictIQ - AI Data Analyst",
    page_icon="🧠",
    layout="wide",
    initial_sidebar_state="expanded"
)

# Load CSS
def local_css(file_name):
    with open(file_name) as f:
        st.markdown(f"<style>{f.read()}</style>", unsafe_allow_html=True)

# Helper for base64 images
def show_b64_img(b64_str):
    st.image(base64.b64decode(b64_str), use_container_width=True)

# Initialize Engine
if 'engine' not in st.session_state:
    # Use environment variable for security
    api_key = os.getenv("GROQ_API_KEY")
    st.session_state.engine = PredictIQEngine(groq_api_key=api_key)
if 'results' not in st.session_state:
    st.session_state.results = None

# App Layout
local_css("style.css")

with st.sidebar:
    st.markdown("<h1 style='color: #818cf8;'>PredictIQ</h1>", unsafe_allow_html=True)
    st.markdown("---")
    uploaded_file = st.file_uploader("Upload Dataset (CSV)", type=["csv"])
    
    if uploaded_file:
        # Reset results if a different file is uploaded
        if 'last_file' not in st.session_state or st.session_state.last_file != uploaded_file.name:
            st.session_state.results_ready = False
            st.session_state.eval_data = None
            st.session_state.last_file = uploaded_file.name
            
        df = pd.read_csv(uploaded_file)
        # Use a consistent variable name 'columns'
        columns = st.session_state.engine.load_data(df)
        
        st.write("### ⚙️ Configuration")
        # Smart Target Detection
        suggested_target = st.session_state.engine.detect_target()
        target_col = st.selectbox(
            "Target Column:", 
            options=columns, 
            index=columns.index(suggested_target)
        )
        
        # Feature Selection
        suggested_drops = st.session_state.engine.get_suggested_drops()
        cols_to_drop = st.multiselect(
            "Exclude Columns:",
            options=[c for c in columns if c != target_col],
            default=[c for c in suggested_drops if c in columns and c != target_col],
            help="PredictIQ identified these as likely IDs or high-cardinality columns."
        )
        
        st.markdown("---")
        if st.button("🚀 Run AI Analysis", use_container_width=True):
            with st.spinner("🧠 Training & Analyzing..."):
                try:
                    # Training with universal parameters
                    st.session_state.engine.train(target_col, columns_to_drop=cols_to_drop)
                    st.session_state.eval_data = st.session_state.engine.evaluate()
                    st.session_state.results_ready = True
                    st.success("Analysis Complete!")
                except Exception as e:
                    st.error(f"Analysis failed: {str(e)}")
                    st.session_state.results_ready = False
                
    st.markdown("---")
    st.info("PredictIQ uses AutoML and Groq LLM to provide deep insights into your datasets.")

# Main Interface
st.markdown("<div class='main-title'>PredictIQ AI Analyst</div>", unsafe_allow_html=True)
st.markdown("<div class='subtitle'>Automated Machine Learning & AI-Powered Data Insights</div>", unsafe_allow_html=True)

if not uploaded_file:
    st.markdown("""
        <div class='glass-card animate-fade-in'>
            <h3>Welcome to PredictIQ</h3>
            <p>Upload a CSV file in the sidebar to begin your automated analysis.</p>
            <ul>
                <li>Automated Feature Engineering</li>
                <li>Problem Type Detection (Classification/Regression)</li>
                <li>Intelligent Feature Selection (ID detection)</li>
                <li>AI-Powered Narrative Insights</li>
            </ul>
        </div>
    """, unsafe_allow_html=True)
else:
    # Data Preview Section
    with st.expander("📊 Dataset Exploration", expanded=True):
        st.write("### Data Preview")
        st.dataframe(df.head(10), use_container_width=True)
        st.info(f"💾 **Stats:** {df.shape[0]} rows | {df.shape[1]} columns")

    if st.session_state.get('results_ready'):
        # RESULTS DASHBOARD
        st.markdown("## 📈 Performance Summary")
        
        col1, col2, col3 = st.columns(3)
        with col1:
            st.markdown(f"""
                <div class='metric-card'>
                    <div class='metric-label'>Best Model</div>
                    <div class='metric-value' style='font-size: 1.5rem;'>{st.session_state.engine.results['best_model']}</div>
                </div>
            """, unsafe_allow_html=True)
        with col2:
            score_label = "Accuracy" if st.session_state.engine.problem_type == "classification" else "R² Score"
            st.markdown(f"""
                <div class='metric-card'>
                    <div class='metric-label'>{score_label}</div>
                    <div class='metric-value'>{st.session_state.engine.best_score:.4f}</div>
                </div>
            """, unsafe_allow_html=True)
        with col3:
            st.markdown(f"""
                <div class='metric-card'>
                    <div class='metric-label'>Problem Type</div>
                    <div class='metric-value'>{st.session_state.engine.problem_type.capitalize()}</div>
                </div>
            """, unsafe_allow_html=True)

        # EDA & Evaluation Tabs
        tab1, tab2, tab3 = st.tabs(["🔍 Exploratory Data Analysis", "🎯 Model Evaluation", "🤖 AI Analyst Report"])

        with tab1:
            st.markdown("### Automated EDA")
            plots = st.session_state.engine.get_eda_plots()
            
            if 'correlation_matrix' in plots:
                st.write("#### Feature Correlation")
                show_b64_img(plots['correlation_matrix'])
            
            st.write("#### Feature Distributions")
            cols = st.columns(2)
            for i, (name, b64) in enumerate(plots.items()):
                if name != 'correlation_matrix':
                    with cols[i % 2]:
                        show_b64_img(b64)

        with tab2:
            st.markdown("### Model Metrics")
            eval_data = st.session_state.eval_data
            
            if st.session_state.engine.problem_type == "classification":
                c1, c2 = st.columns([1, 1])
                with c1:
                    st.write("#### Classification Report")
                    report_df = pd.DataFrame(eval_data['report']).transpose()
                    st.dataframe(report_df.style.highlight_max(axis=0))
                with c2:
                    st.write("#### Confusion Matrix")
                    show_b64_img(eval_data['confusion_matrix'])
            else:
                st.write("#### Regression Metrics")
                metrics = eval_data['metrics']
                m_col1, m_col2, m_col3 = st.columns(3)
                m_col1.metric("R² Score", f"{metrics['R2']:.4f}")
                m_col2.metric("MAPE", f"{metrics['MAPE']:.4f}")
                m_col3.metric("RMSE", f"{metrics['RMSE']:.4f}")

        with tab3:
            st.markdown("### AI-Generated Insights")
            st.markdown(f"""
                <div class='glass-card' style='border-left: 5px solid #a855f7;'>
                    {st.session_state.eval_data['ai_insights']}
                </div>
            """, unsafe_allow_html=True)
            
            # Download Report
            report_text = f"PredictIQ Analysis Report\n" + "="*25 + "\n"
            report_text += f"Best Model: {st.session_state.engine.results['best_model']}\n"
            report_text += f"Score: {st.session_state.engine.best_score:.4f}\n\n"
            report_text += "AI ANALYST INSIGHTS:\n" + st.session_state.eval_data['ai_insights']
            
            st.download_button(
                label="📥 Download Analysis Report",
                data=report_text,
                file_name="predictiq_report.txt",
                mime="text/plain"
            )

        # Download Model
        model_path = st.session_state.engine.save_model()
        with open(model_path, "rb") as f:
            st.download_button(
                label="💾 Download Best Model (.pkl)",
                data=f,
                file_name="best_model.pkl",
                mime="application/octet-stream"
            )