Upload folder using huggingface_hub

Dockerfile

@@ -0,0 +1,28 @@
+# Dockerfile
+
+# Use a specific Python base image for stability and size optimization
+FROM python:3.12-slim
+
+# Set the working directory inside the container
+WORKDIR /app
+
+# Copy the requirements file and install dependencies
+# Use --no-cache-dir to keep the image size small
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy the Flask application and the serialized model
+COPY app.py .
+COPY tuned_xgb_sales_forecaster.pkl .
+
+# Expose the port the Flask app will run on
+EXPOSE 7860
+
+# Command to Start the Application (Gunicorn)
+# This is the crucial part, borrowing from your colleague's working model:
+# - `-w 4`: 4 worker processes for concurrency
+# - `-b 0.0.0.0:7860`: Binds to the required port and all interfaces
+# - `app:app`: The application target. It means: 
+#   - look in file 'app.py' (the first 'app')
+#   - for the Flask instance named 'app' (the second 'app')
+CMD ["gunicorn", "-w", "4", "-b", "0.0.0.0:7860", "app:app"]

app.py

@@ -0,0 +1,71 @@
+# app.py
+
+# Import necessary libraries
+import joblib
+import numpy as np
+import pandas as pd
+from flask import Flask, request, jsonify
+
+# Initialize the Flask application
+# Using __name__ is essential for Flask to correctly find its resources.
+app = Flask(__name__) 
+app.name = "SuperKart Sales Forecaster" # Explicitly set the name property for logging
+
+# Define the model filename
+# NOTE: Ensure the path 'backend_files/tuned_xgb_sales_forecaster.pkl' is correct
+# relative to where app.py is run.
+MODEL_FILE = 'backend_files/tuned_xgb_sales_forecaster.pkl'
+
+# Define the 10 feature columns expected by the model pipeline
+FEATURE_COLS = [
+    'Product_Weight', 'Product_Sugar_Content', 'Product_Allocated_Area',
+    'Product_Type', 'Product_MRP', 'Store_Size',
+    'Store_Location_City_Type', 'Store_Type', 'Store_Age',
+    'Product_Category_Simplified'
+]
+
+# --- Load the Model Pipeline (Global scope, loads once on startup) ---
+try:
+    model_pipeline = joblib.load(MODEL_FILE)
+    print("Model loaded successfully.")
+except Exception as e:
+    print(f"CRITICAL ERROR: Model not found at {MODEL_FILE}: {e}. Check Dockerfile.")
+    model_pipeline = None
+
+@app.route('/predict', methods=['POST'])
+def predict_sales():
+    if model_pipeline is None:
+        return jsonify({'error': 'Server setup error: Model not loaded.'}), 500
+
+    try:
+        # force=True is often necessary when content-type headers are not set perfectly
+        data = request.get_json(force=True)
+        
+        # Ensure data is a list of dictionaries, even if we only send one
+        if not isinstance(data, list) or not data:
+             return jsonify({'error': 'Input data must be a non-empty list of prediction objects.'}), 400
+
+        # Ensure input data matches the feature columns
+        input_df = pd.DataFrame(data, columns=FEATURE_COLS)
+
+        # Prediction on log scale
+        log_prediction = model_pipeline.predict(input_df)
+
+        # Inverse transformation: sales = exp(y) - 1
+        prediction_original_scale = np.expm1(log_prediction)
+
+        response = {
+            'status': 'success',
+            # We only expect one prediction, so take the first element [0]
+            'predicted_sales_revenue': round(prediction_original_scale[0], 2)
+        }
+        return jsonify(response)
+
+    except Exception as e:
+        # Log the detailed exception on the server side for debugging
+        print(f"Prediction failed with exception: {e}") 
+        return jsonify({'error': f'Prediction logic failed: {str(e)}'}), 400
+
+if __name__ == '__main__':
+    # This runs the server inside the Docker container
+    app.run(host='0.0.0.0', port=5000)

requirements.txt

@@ -0,0 +1,6 @@
+Flask==3.0.3
+pandas==2.2.2
+numpy==1.26.4
+scikit-learn==1.6.1
+joblib==1.4.2
+xgboost==2.1.4

tuned_xgb_sales_forecaster.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64e94ad83c2cb59138fd0addc972f5484caad0c21455161edb6465fafe8f24ae
+size 612275