The Python SQL ecosystem is IMO pretty bad, and the root cause seems to be that DB-API (PEP 249) is really quite awful.
As my least favorite example, DB-API has two transaction modes, both of which are unnecessarily error prone: autocommit on, and autocommit off. Autocommit on makes transactions useless (intentionally), but it still allows commit() and rollback(), so one can write code that thinks transactions are enabled, have it appear to work, and end up with all manner of corruption.
Autocommit off has a bad interface. The begin() function is entirely missing, so you can’t pass it arguments. Worse, since there is no explicit start of a transaction, the library can’t catch misuse of a transaction.
As a result, every SQL library for Python of any reasonable quality implements its own special API, and nothing is particularly interoperable.
This is all sort of fine if the database is only accessed through a higher level library like SQLAlchemy. But sometimes a developer just wants to execute raw SQL against a database, and the situation in Python is not very nice.