app.py

import gradio as gr
import pandas as pd
import numpy as np
from sklearn.preprocessing import LabelEncoder
from sklearn.ensemble import RandomForestRegressor
import matplotlib.pyplot as plt
import io

# --------------------------
# Global variables
# --------------------------
DATA = None
ENCODERS = {}
MODEL = None

# --------------------------
# Load and preprocess dataset
# --------------------------
def load_dataset(file_obj):
    global DATA, ENCODERS
    try:
        df = pd.read_csv(file_obj.name)
    except:
        df = pd.read_csv(file_obj)
    preview, msg = preprocess_data(df)
    return preview, msg

def preprocess_data(df):
    global DATA, ENCENCODERS
    required_cols = [
        "date", "Usage_kWh",
        "Lagging_Current_Reactive.Power_kVarh",
        "Leading_Current_Reactive_Power_kVarh",
        "CO2(tCO2)",
        "Lagging_Current_Power_Factor",
        "Leading_Current_Power_Factor",
        "NSM",
        "WeekStatus",
        "Day_of_week",
        "Load_Type"
    ]
    missing = [c for c in required_cols if c not in df.columns]
    if missing:
        return pd.DataFrame(), f"❌ Missing columns: {missing}"

    df['date'] = pd.to_datetime(df['date'], errors='coerce')
    df['month'] = df['date'].dt.month

    cat_cols = ["WeekStatus", "Day_of_week", "Load_Type"]
    global ENCODERS
    ENCODERS = {}
    for col in cat_cols:
        enc = LabelEncoder()
        df[col] = enc.fit_transform(df[col].astype(str))
        ENCODERS[col] = enc

    global DATA
    DATA = df

    return df.head(10), "✅ Dataset uploaded and processed successfully!"

# --------------------------
# Train model
# --------------------------
def train_model():
    global DATA, MODEL
    if DATA is None:
        return "❌ Upload dataset first!"

    features = [
        "month",
        "Lagging_Current_Reactive.Power_kVarh",
        "Leading_Current_Reactive_Power_kVarh",
        "CO2(tCO2)",
        "Lagging_Current_Power_Factor",
        "Leading_Current_Power_Factor",
        "NSM",
        "WeekStatus",
        "Day_of_week",
        "Load_Type"
    ]
    X = DATA[features]
    y = DATA["Usage_kWh"]
    MODEL = RandomForestRegressor(n_estimators=100, random_state=42)
    MODEL.fit(X, y)
    return "✅ Model trained successfully!"

# --------------------------
# Generate prediction
# --------------------------
def generate_prediction(load, week, day, co2, lag, lead, usage, nsm):
    global MODEL, ENCODERS
    if MODEL is None:
        return "❌ Model not trained yet!"

    try:
        load_enc = ENCODERS["Load_Type"].transform([load])[0]
        week_enc = ENCODERS["WeekStatus"].transform([week])[0]
        day_enc = ENCODERS["Day_of_week"].transform([day])[0]

        X_input = np.array([[lag, lead, co2, usage, nsm, load_enc, week_enc, day_enc, 0, 0]])
        X_input = X_input[:, :10]
        pred = MODEL.predict(X_input)[0]
        return round(pred, 2)
    except Exception as e:
        return f"❌ Prediction error: {e}"

# --------------------------
# Generate plot
# --------------------------
def generate_plot():
    global DATA
    if DATA is None or DATA.empty:
        return None
    fig, ax = plt.subplots(figsize=(10,5))
    ax.plot(DATA['date'], DATA['Usage_kWh'], color='#2563EB')
    ax.set_title("Energy Usage Over Time", color='#0F172A', fontsize=16)
    ax.set_xlabel("Date")
    ax.set_ylabel("Usage (kWh)")
    ax.grid(True, linestyle='--', alpha=0.5)

    buf = io.BytesIO()
    fig.savefig(buf, format='png', bbox_inches='tight')
    buf.seek(0)
    plt.close(fig)
    return buf

# --------------------------
# Chatbot function
# --------------------------
def on_user_message(msg, load, week, day, co2, lag, lead, usage, nsm):
    if msg.strip() == "":
        return [], msg

    advice = []

    if DATA is not None:
        mean_usage = DATA['Usage_kWh'].mean()
        if usage > mean_usage:
            advice.append("⚠️ Your usage is above average. Check machinery.")

    if lag > 50:
        advice.append("Consider power factor correction for lagging kVarh.")
    if lead > 50:
        advice.append("Check capacitive reactive power usage.")

    if DATA is not None:
        co2_75 = np.percentile(DATA['CO2(tCO2)'], 75)
        if co2 > co2_75:
            advice.append("⚠️ High CO₂ detected. Improve ventilation or reduce production temporarily.")

    if not advice:
        advice.append("✅ All systems normal. No major issues detected.")

    # Return messages as dictionaries (Gradio v3.43+)
    return [
        {"role": "user", "content": msg},
        {"role": "assistant", "content": "\n".join(advice)}
    ], ""

# --------------------------
# Gradio UI
# --------------------------
with gr.Blocks() as app:

    gr.Markdown("## ⚡ Smart Factory Energy Advisor")
    gr.Markdown("Analyze energy usage, detect inefficiencies, and receive actionable recommendations.")

    # Dataset upload
    file = gr.File(label="Upload CSV")
    upload_btn = gr.Button("Upload Dataset")
    preview = gr.Dataframe(label="Dataset Preview (First 10 Rows)", interactive=False)
    status_box = gr.Textbox(label="Status Message", interactive=False)

    upload_btn.click(
        fn=load_dataset,
        inputs=file,
        outputs=[preview, status_box]
    )

    # User input panel
    gr.Markdown("### ⚙️ User Input Panel")
    with gr.Row():
        with gr.Column(scale=1):
            load = gr.Dropdown(["Light","Medium","Maximum"], label="Load Type")
            week = gr.Dropdown(["Weekday","Weekend"], label="Week Status")
            day = gr.Dropdown(
                ["Sunday","Monday","Tuesday","Wednesday","Thursday","Friday","Saturday"], 
                label="Day of Week"
            )
            co2 = gr.Number(label="CO₂ Level (tCO2)")
            lag = gr.Number(label="Lagging Reactive Power (kVarh)")
            lead = gr.Number(label="Leading Reactive Power (kVarh)")
            usage = gr.Number(label="Current Usage (kWh)")
            nsm = gr.Number(label="NSM (Seconds from Midnight)")

    # Chatbot
    gr.Markdown("### 🤖 Chatbot")
    chatbot = gr.Chatbot()
    msg = gr.Textbox(label="Ask a question or request advice/prediction")
    send_btn = gr.Button("Send")
    send_btn.click(
        fn=on_user_message,
        inputs=[msg, load, week, day, co2, lag, lead, usage, nsm],
        outputs=[chatbot, msg]
    )

    # Train & Predict
    gr.Markdown("### 🔮 Energy Prediction")
    train_btn = gr.Button("Train Model")
    train_out = gr.Textbox(label="Training Status", interactive=False)
    train_btn.click(fn=train_model, inputs=[], outputs=train_out)

    pred_btn = gr.Button("Predict Energy Usage")
    pred_out = gr.Textbox(label="Predicted Usage (kWh)", interactive=False)
    pred_btn.click(
        fn=generate_prediction,
        inputs=[load, week, day, co2, lag, lead, usage, nsm],
        outputs=pred_out
    )

    # Plot chart
    gr.Markdown("### 📈 Consumption Chart")
    plot_btn = gr.Button("Show Consumption Chart")
    plot_img = gr.Image()
    plot_btn.click(fn=generate_plot, inputs=[], outputs=plot_img)

app.launch()