Add Jared read-only token OHLCV ingestion and manual contract lookup

core/data.py

@@ -509,11 +509,14 @@ def preprocess_data(df, features, target, window_size, horizon):
             raise ValueError("DataFrame index must be a DatetimeIndex for preprocessing.")
 
         # Filter features to only include those present in the DataFrame columns
-        updated_feature_cols = [f for f in features if f in df.columns]
-        if not updated_feature_cols:
-            raise ValueError("No valid features found in DataFrame after indicator calculation.")
-
-        full_features = updated_feature_cols + [target]
+        # Never include the target twice. The fallback indicator path may
+        # return ['value']; duplicating it here makes target_scaler fit two
+        # columns and breaks multi-step inverse scaling.
+        updated_feature_cols = list(dict.fromkeys(
+            f for f in features if f in df.columns and f != target
+        ))
+        model_feature_cols = updated_feature_cols or [target]
+        full_features = list(dict.fromkeys(model_feature_cols + [target]))
         data = df[full_features].copy()
         data.dropna(inplace=True)
 
@@ -523,11 +526,11 @@ def preprocess_data(df, features, target, window_size, horizon):
         feature_scaler = MinMaxScaler()
         target_scaler = MinMaxScaler()
 
-        data_features_scaled = feature_scaler.fit_transform(data[updated_feature_cols])
+        data_features_scaled = feature_scaler.fit_transform(data[model_feature_cols])
         data_target_scaled = target_scaler.fit_transform(data[[target]])
 
         full_scaled = np.hstack((data_features_scaled, data_target_scaled))
-        target_idx = len(updated_feature_cols)
+        target_idx = len(model_feature_cols)
 
         X, y = [], []
         for i in range(len(full_scaled) - window_size - horizon + 1):
@@ -541,7 +544,7 @@ def preprocess_data(df, features, target, window_size, horizon):
             raise ValueError(f"Insufficient data after preprocessing. Data length: {len(full_scaled)}, window_size: {window_size}, horizon: {horizon}")
 
         print_log(f"Preprocessed data: X.shape={X.shape}, y.shape={y.shape}, Final features: {full_features}, Target idx: {target_idx}")
-        return X, y, feature_scaler, target_scaler, full_features, target_idx, None, updated_feature_cols
+        return X, y, feature_scaler, target_scaler, full_features, target_idx, None, model_feature_cols
     except Exception as e:
         print_log(f"Preprocessing error: {str(e)}", 'ERROR')
         raise ValueError(f"Preprocessing failed: {str(e)}")