app.py

import joblib
import pandas as pd
from flask import Flask, request, jsonify

# Initialize Flask app
sales_forecast_api = Flask("SuperKart Sales Forecast API")

# Load trained model (includes preprocessing pipeline)
model = joblib.load("best_random_forest.pkl")

# Home route
@sales_forecast_api.get('/')
def home():
    return "Welcome to the SuperKart Sales Forecast API!"

# Predict for a single input
@sales_forecast_api.post('/v1/predict')
def predict_sales():
    data = request.get_json()

    # Convert JSON into DataFrame (1 row)
    input_df = pd.DataFrame([data])

    # Model prediction
    prediction = model.predict(input_df).tolist()[0]

    return jsonify({"Predicted_Sales_Total": prediction})

# Predict for a batch (CSV upload)
@sales_forecast_api.post('/v1/predict_batch')
def predict_sales_batch():
    # Read uploaded CSV file
    file = request.files['file']
    input_data = pd.read_csv(file)

    # Check for optional Product_Id column
    if "Product_Id" in input_data.columns:
        ids = input_data["Product_Id"].astype(str)
        input_data = input_data.drop(columns=["Product_Id"])
    else:
        ids = input_data.index.astype(str)  # fallback to row indices

    # Make predictions
    preds = model.predict(input_data).tolist()

    # Map Product_Id (or row index) → prediction
    results = dict(zip(ids, preds))

    return jsonify(results)

# Run app
if __name__ == "__main__":
    sales_forecast_api.run(debug=True)