Issue with seed data in Alembic

Question

I have been using SQLAlchemy for year now. And I use Alembic for migrating changes.

I am also using alembic to seed data. Take an example of the authorization seeds.

def upgrade():
    op.bulk_insert(Role.__table__, ROLE_LIST)
    op.bulk_insert(Permission.__table__, PERMISSION_LIST)
    op.bulk_insert(RolePermission.__table__, ROLE_PERMISSION_LIST)

def downgrade():
    [op.execute(a_table.__table__.delete()) for a_table in [RolePermission, Permission, Role]]

This works perfect if we assume there are going to be no changes in my data set. (permissions, roles, etc in my case)

ROLE_LIST = [
    dict(id=generate_business_id(), name='customer', default=1),
    dict(id=generate_business_id(), name='admin', default=0), 
]
PERMISSION_LIST = [
    dict(id=generate_business_id(), name='profile'),
    dict(id=generate_business_id(), name='delete_user'),
    dict(id=generate_business_id(), name='make_payment'),
]

But it's not the case. I am gonna have new roles and permissions and role_permission mappings. And alembic upgrade head will not migrate new data because my head has gone far ahead of the seed migration.

I could have the data in the respective revision only, but I am also using ROLE_LIST and PERMISSION_LIST in my business logic.

Is there any way I can do replace query in migrations? Is there any better option?