Update app.py

app.py

@@ -102,45 +102,59 @@ def predict(*feature_values):
             proba_dict: dict(label -> probability), sorted desc, top-N shown by gr.Label
             message:    Markdown summary with predicted label + confidence
     """
-    # Map UI inputs to a dict matching the model's feature columns
-    input_data = {}
-    for col, val in zip(FEATURE_COLS, feature_values[:len(FEATURE_COLS)]):
-        try:
-            # Try numeric first (keeps sliders/numbers numeric)
-            input_data[col] = float(val) if val != "" else 0.0
-        except:
-            # Otherwise leave as string (for categorical columns)
-            input_data[col] = val
-    
-    # Build a DataFrame row for inference
-    X = pd.DataFrame([input_data])
-    
-    # Predicted label (or regression value)
-    pred = PREDICTOR.predict(X)
-    pred_value = pred.iloc[0]
-    
-    # Class probabilities (if classifier). If regression, synthesize 100% on prediction.
     try:
-        proba_df = PREDICTOR.predict_proba(X)
-        if isinstance(proba_df, pd.Series):
-            # Normalize to DataFrame shape if AG returns a Series
-            proba_df = proba_df.to_frame().T
+        # Map UI inputs to a dict matching the model's feature columns
+        input_data = {}
+        for col, val in zip(FEATURE_COLS, feature_values[:len(FEATURE_COLS)]):
+            try:
+                # Try numeric first (keeps sliders/numbers numeric)
+                input_data[col] = float(val) if val != "" else 0.0
+            except:
+                # Otherwise leave as string (for categorical columns)
+                input_data[col] = val
         
-        proba_dict = {}
-        for col in proba_df.columns:
-            proba_dict[str(col)] = float(proba_df[col].iloc[0])
+        print(f"Input data: {input_data}")
         
-        # Sort highest to lowest
-        proba_dict = dict(sorted(proba_dict.items(), key=lambda x: x[1], reverse=True))
-    except:
-        # Regression or unsupported proba: show pseudo-confidence
-        proba_dict = {str(pred_value): 1.0}
-    
-    # Human-readable summary (confidence = max probability * 100)
-    confidence = max(proba_dict.values()) * 100 if proba_dict else 100
-    message = f"**Prediction:** {pred_value}\n**Confidence:** {confidence:.2f}%"
-    
-    return proba_dict, message
+        # Build a DataFrame row for inference
+        X = pd.DataFrame([input_data])
+        print(f"DataFrame shape: {X.shape}")
+        print(f"DataFrame columns: {X.columns.tolist()}")
+        
+        # Predicted label (or regression value)
+        pred = PREDICTOR.predict(X)
+        pred_value = pred.iloc[0]
+        print(f"Prediction: {pred_value}")
+        
+        # Class probabilities (if classifier). If regression, synthesize 100% on prediction.
+        try:
+            proba_df = PREDICTOR.predict_proba(X)
+            if isinstance(proba_df, pd.Series):
+                # Normalize to DataFrame shape if AG returns a Series
+                proba_df = proba_df.to_frame().T
+            
+            proba_dict = {}
+            for col in proba_df.columns:
+                proba_dict[str(col)] = float(proba_df[col].iloc[0])
+            
+            # Sort highest to lowest
+            proba_dict = dict(sorted(proba_dict.items(), key=lambda x: x[1], reverse=True))
+        except Exception as e:
+            print(f"Error getting probabilities: {e}")
+            # Regression or unsupported proba: show pseudo-confidence
+            proba_dict = {str(pred_value): 1.0}
+        
+        # Human-readable summary (confidence = max probability * 100)
+        confidence = max(proba_dict.values()) * 100 if proba_dict else 100
+        message = f"**Prediction:** {pred_value}\n**Confidence:** {confidence:.2f}%"
+        
+        return proba_dict, message
+        
+    except Exception as e:
+        error_msg = f"**Error:** {str(e)}\n\nPlease check the logs for details."
+        print(f"Prediction error: {e}")
+        import traceback
+        traceback.print_exc()
+        return {}, error_msg
 
 
 # ============================================================================