How to revert keras sequential model to a state previous to build?

Question

I want to reuse the same model architecture but with different datasets, that is, to programmatically change the input layer to a different shape, and reset model parameters