Upload folder using huggingface_hub

app.py

@@ -10,15 +10,13 @@ import os
 
 # --- Configuration ---
 MODEL_PATH = "final_xgboost_pipeline.pkl"
-APP_NAME = "SuperKart Sales Predictor"
 
 # --- Initialize Flask App ---
 app = Flask("SuperKart Sales Predictor")
 
 # --- Global Model Loading ---
 try:
-    # model = joblib.load(MODEL_PATH)
-    model = MODEL_PATH
+    model = joblib.load(MODEL_PATH)
     print("Model loaded successfully.") # Add logging
 except Exception as e:
     print(f"Error loading model: {e}") # Log error
@@ -33,12 +31,12 @@ def health_check():
     print("Home route accessed.") # Add logging
     return jsonify({
         "status": "OK",
-        "service": APP_NAME,
         "model_location": MODEL_PATH,
         "model_status": "Loaded" if model else "Error: See logs",
         "api_version": "No Versioning"
     })
 
+
 @app.route("/predict", methods=['POST']) # The simple, unversioned route
 def predict_sales():
     """
@@ -50,21 +48,25 @@ def predict_sales():
         return jsonify({'error': 'Prediction service unavailable. Model failed to load.'}), 503
 
     try:
-        data = request.get_json(force=True)
+        # 1. REMOVE force=True to enforce correct MIME type header
+        data = request.get_json() 
         print(f"[{APP_NAME}] Received prediction request with data: {data}")
 
         if not data:
-            return jsonify({'error': 'Invalid or empty JSON data provided.'}), 400
+            return jsonify({'error': 'Invalid or empty JSON data provided. Ensure Content-Type is application/json.'}), 400
 
         # --- Data Transformation ---
         try:
             # Create a one-row DataFrame
             input_df = pd.DataFrame([data])
-            # Ensure Store_Establishment_Year is treated as categorical
-            input_df['Store_Establishment_Year'] = input_df['Store_Establishment_Year'].astype(str)
-            print(f"[{APP_NAME}] Input data has been transformed successfully.")
-            output = "Input data has been transformed successfully."
-
+            
+            # 2. REMOVE MANUAL TYPE CONVERSION
+            # A well-built ML pipeline should handle all feature transformations internally.
+            # input_df['Store_Establishment_Year'] = input_df['Store_Establishment_Year'].astype(str)
+            
+            print(f"[{APP_NAME}] Input data prepared for pipeline.")
+            # output variable removed as it was only used for the previous error
+            
         except KeyError as ke:
             print(f"[{APP_NAME}] Missing required feature: {ke}")
             return jsonify({'error': f'Missing required feature: {ke}. Please check all features are present.'}), 400
@@ -81,14 +83,15 @@ def predict_sales():
         if isinstance(prediction, (np.float32, np.float64)):
             predicted_sales = float(prediction)
         else:
-            predicted_sales = prediction
+            predicted_sales = prediction # Should ideally be a float
 
         print(f"[{APP_NAME}] Prediction result: {predicted_sales:.2f}")
 
         # --- Response ---
+        # 3. FIX: Return the actual predicted value instead of the string message
         return jsonify({
             'status': 'success',
-            'predicted_sales_total': output
+            'predicted_sales': predicted_sales # Renamed key to 'predicted_sales' for clarity
         })
 
     except Exception as e:
@@ -99,7 +102,7 @@ def predict_sales():
             'error': 'Internal Server Error during prediction',
             'details': str(e)
         }), 500
-
+        
 
 # --- Local Runner (Optional: Comment out for production WSGI) ---
 if __name__ == '__main__':