app.py

import json
import os
import joblib
import pandas as pd
import streamlit as st
from backend.train_model import train_model

MODEL_DIR = "models"
MODEL_FILE = "my_model.pkl"
MODEL_PATH = os.path.join(MODEL_DIR, MODEL_FILE)

REPORTS_DIR = "reports"
PLOTS_DIR = os.path.join(REPORTS_DIR, "plots")

FEATURES = [
    "Pregnancies", "Glucose", "BloodPressure", "SkinThickness",
    "Insulin", "BMI", "DiabetesPedigreeFunction", "Age"
]

st.set_page_config(page_title="Diabetes Prediction Dashboard", layout="wide")
st.title("🩺 Diabetes Prediction Dashboard")

# Sidebar navigation
st.sidebar.header("Navigation")
page = st.sidebar.radio("Go to", ["Predict", "Batch Predict", "Reports & Plots"])

model = None
if os.path.exists(MODEL_PATH):
    model = joblib.load(MODEL_PATH)

# ------------------ Train button ------------------
st.subheader("Train & Predict Diabetes Model")

if st.button("Train Model"):
    with st.spinner("Training in progress... this may take a while ⏳"):
        model = train_model()
        joblib.dump(model, MODEL_PATH)
    st.success(f"✅ Model trained and saved to `{MODEL_PATH}`")

# ------------------ Predict single ------------------
def predict_df(df: pd.DataFrame):
    if model is None:
        st.error("⚠️ Model not loaded. Train first.")
        return None
    missing = [c for c in FEATURES if c not in df.columns]
    if missing:
        st.error(f"Missing columns: {missing}")
        return None
    return model.predict(df[FEATURES])

if page == "Predict":
    st.subheader("🔹 Single Prediction")
    cols = st.columns(4)
    values = {}
    ranges = {
        "Pregnancies": (0, 20, 1), "Glucose": (0, 220, 120),
        "BloodPressure": (0, 150, 70), "SkinThickness": (0, 100, 20),
        "Insulin": (0, 900, 80), "BMI": (0.0, 70.0, 25.0),
        "DiabetesPedigreeFunction": (0.0, 3.0, 0.5), "Age": (0, 120, 30)
    }
    for i, f in enumerate(FEATURES):
        with cols[i % 4]:
            lo, hi, default = ranges[f]
            if isinstance(default, float):
                values[f] = st.number_input(f, lo, hi, float(default))
            else:
                values[f] = st.number_input(f, int(lo), int(hi), int(default))
    if st.button("Predict"):
        row = pd.DataFrame([values])
        pred = predict_df(row)
        if pred is not None:
            st.success("✅ Diabetic" if int(pred[0]) == 1 else "🟢 Not Diabetic")

# ------------------ Batch predict ------------------
elif page == "Batch Predict":
    st.subheader("📂 Batch Prediction (Upload CSV)")
    st.caption("CSV must include columns: " + ", ".join(FEATURES))
    file = st.file_uploader("Upload CSV", type=["csv"])
    if file is not None:
        df = pd.read_csv(file)
        st.write("Preview of uploaded data:")
        st.dataframe(df.head())
        preds = predict_df(df)
        if preds is not None:
            out = df.copy()
            out["Prediction"] = preds
            st.success(f"Predicted {len(out)} rows")
            st.dataframe(out.head())
            st.download_button(
                "⬇️ Download predictions",
                data=out.to_csv(index=False).encode('utf-8'),
                file_name="predictions.csv",
                mime="text/csv"
            )

# ------------------ Reports & plots ------------------
elif page == "Reports & Plots":
    st.subheader("📊 Model Comparison & Diagnostics")

    # Load CSV
    perf_csv = os.path.join(REPORTS_DIR, "model_comparison.csv")
    perf_json = os.path.join(REPORTS_DIR, "model_comparison.json")

    if os.path.exists(perf_csv):
        df_perf = pd.read_csv(perf_csv)
        st.dataframe(df_perf)

        # Highlight best model based on F1
        best_model = df_perf.loc[df_perf['F1'].idxmax()]
        st.success(f"Best Model: {best_model['Model']} ✅ F1: {best_model['F1']} Accuracy: {best_model['Accuracy']}")

        # Bar chart
        st.bar_chart(df_perf.set_index('Model')[['Accuracy', 'F1']])
    else:
        st.info("No model comparison CSV found. Train models first!")

    # Optionally show JSON
    # if os.path.exists(perf_json):
    #     with open(perf_json, "r") as f:
    #         json_data = json.load(f)
    #     st.json(json_data)

    # Display plots
    plot_files = [
        ("Accuracy (bar)", "model_accuracy.png"),
        ("F1 (bar)", "model_f1.png"),
        ("Confusion Matrix (best)", "confusion_matrix.png"),
        ("ROC (best)", "roc_curve.png"),
        ("Variance (before/after)", "variance_comparison.png"),
        ("LR Loss vs Iterations", "logreg_loss_curves.png"),
        ("LR Accuracy vs Iterations", "logreg_accuracy_curves.png"),
    ]
    rows = st.columns(2)
    for i, (title, fname) in enumerate(plot_files):
        p = os.path.join(PLOTS_DIR, fname)
        if os.path.exists(p):
            with rows[i % 2]:
                st.markdown(f"**{title}**")
                st.image(p)
        else:
            st.info(f"{fname} not available yet.")