Update model_pipeline.py

model_pipeline.py

@@ -261,6 +261,33 @@ class Predictor:
             down = short([n for n in neg_pairs if n[1] < 0])
             return (f"Classe **{final_class}** (spiegazione su P(y≥{k_thr})): "
                     f"in alto {up}; in basso {down}. Valore atteso {yhat:.1f}.")
+            
+    def predict_class_fast(self, payload: dict):
+        """Like predict_dict ma senza SHAP: restituisce solo classe, p100, probs ordinali e valore atteso."""
+        raw = {k: payload.get(k, None) for k in FEATURE_MAP.values()}
+        df_row_raw = pd.DataFrame([raw])
+
+        te_oh = self.preprocess_apply(df_row_raw)
+        X_df = te_oh.reindex(columns=self.feat_cols_full, fill_value=0.0)
+        X = X_df.values
+
+        p100 = float(self.stage1_final.predict_proba(X)[:,1][0])
+        prob_ord = self.stage2_final.predict_proba(X)[0]
+        prob_ord = prob_ord / (prob_ord.sum() or 1.0)
+        yhat = 100.0*p100 + (1.0-p100)*float((prob_ord @ MIDPOINTS))
+
+        if p100 >= P100_THR_AUTO:
+            final_class = "100%"
+        else:
+            k = int(np.argmax(prob_ord))
+            final_class = LABELS[k]
+
+        return {
+            "class": final_class,
+            "p100": p100,
+            "ordinal_probs": {LABELS[i]: float(prob_ord[i]) for i in range(len(LABELS))},
+            "expected_value": float(yhat)
+        }
 
     def predict_dict(self, payload: dict, include_neg: bool=False):
         rid = str(uuid.uuid4())