[QUESTION] Why Gradient Boost only uses Decision Trees

I wonder after reading Chapter 7 that in Ada Boost Ensemble Algorithm multiple types of base learning algorithms can be used ex: decision trees, svm, logistic regression, etc. But in the implementation of Gradient Boosting algorithm in sklearn the weak learners are fixed to be type decision trees, why don't it give freedom and equal rights to other algorithms. There may be a reason to do so that's my question anyway why?