Description of Problem:
If the user wants to add his contributions computed with a model other than those compatible with shapash, he must add a fake model to the model argument that is compatible with shapash, for example :

fake_model = catboost.CatBoostRegressor(verbose=False).fit(catboost.Pool(x_train, y_train))

xpl = SmartExplainer(model=fake_model)

xpl.compile(x=x, contributions=user_contributions)

Overview of the Solution:
Make the model optional, or raise the contributions parameter in the init, or whatever.