app.py

import pandas as pd
from flask import Flask, request, jsonify
import joblib
import numpy as np
import logging
import sys

# -----------------------
# Setup logger (stdout captured by HF Spaces)
# -----------------------
logger = logging.getLogger("capacity_logger")
logger.setLevel(logging.INFO)
handler = logging.StreamHandler(sys.stdout)  # important for HF Spaces
handler.setLevel(logging.INFO)
formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s')
handler.setFormatter(formatter)
logger.addHandler(handler)

# -----------------------
# Initialize Flask app
# -----------------------
app = Flask("Store Capacity Predictor")

# Load the trained pipeline
pipeline = joblib.load("catbooster_model_v1_0.joblib")

# -----------------------
# Home route
# -----------------------
@app.get('/')
def home():
    return "Welcome to the Store Capacity Prediction API"

# -----------------------
# Single prediction
# -----------------------
@app.post('/v1/predict')
def predict_capacity():
    try:
        sales_data = request.get_json()
        input_data = pd.DataFrame([sales_data])
        input_data["Date"] = pd.to_datetime(input_data["Date"])

        logger.info("Single prediction input:\n%s", input_data)

        # Predict
        prediction = pipeline.predict(input_data).tolist()[0]

        # Sanitize prediction
        if not np.isfinite(prediction):
            logger.warning("Single prediction invalid (%s), replacing with 0", prediction)
            prediction = 0
        else:
            prediction = int(np.clip(prediction, 0, 10000))

        logger.info("Single prediction output: %s", prediction)

        return jsonify({'Predicted_Capacity': prediction})

    except Exception as e:
        logger.error("Error in single prediction: %s", e, exc_info=True)
        return jsonify({'error': str(e)}), 400

# -----------------------
# Batch prediction
# -----------------------
@app.post('/v1/predict_batch')
def predict_capacity_batch():
    try:
        # Get the uploaded CSV file from the request
        file = request.files['file']

        # Read the file into a DataFrame
        input_data = pd.read_csv(file)
        #input_data = pd.DataFrame(data_list)
        input_data["Date"] = pd.to_datetime(input_data["Date"])
        input_data["SpecialEvent"] = input_data["SpecialEvent"].fillna("").astype(str)
      
        logger.info("Batch input shape: %s", input_data.shape)
        logger.info("Batch input preview:\n%s", input_data.head())

        predictions = pipeline.predict(input_data).tolist()
        clean_predictions = []

        for idx, p in enumerate(predictions):
            if not np.isfinite(p):
                logger.warning("Row %d prediction invalid (%s), replacing with 0", idx, p)
                clean_predictions.append(0)
            else:
                clean_predictions.append(int(np.clip(p, 0, 10000)))

        logger.info("Batch predictions: %s", clean_predictions)

        output_df = pd.DataFrame({
            "BU": input_data["BU"],
            "Date": input_data["Date"],
            "Store": input_data["Store"],
            "LocationType": input_data["LocationType"],
            "Slot": input_data["Slot"],
            "Predicted_Capacity": clean_predictions
        })

        return output_df.to_html(index=False)

    except Exception as e:
        logger.error("Error in batch prediction: %s", e, exc_info=True)
        return jsonify({"error": str(e)}), 400

# -----------------------
# Run Flask app
# -----------------------
if __name__ == '__main__':
    app.run(debug=True)