- currently all our database code goes through this service - theoretically users should be able to have one relational DB for their columnar data and be able to swap out the vector db they want to use