There are several wrappers that provide improved or simplified interfaces to SQL databases. Some of these might be referred to as object relational mappers, or ORM in this list -- these create Pythonic objects out of database rows. Others may only help generate SQL, or provide simple mapping support.
DbObj: ORM
[http://opensource.theopalgroup.com/ db_row]: SQL result wrapper
[http://forgetsql.sourceforge.net/ forgetSQL]: ORM
[http://www.livinglogic.de/Python/sql/ ll.sql]: SQL generator
PgResultSet: built-in facility for column access by name for pyPgSQL and PySQLite
MiddleKit: ORM
[http://modeling.sourceforge.net/ Modeling Object-Relational Bridge]: ORM and schema design with Zope integration
[http://www.tux4web.de/orm/ Object Relational Membrame]: ORM
[http://pdo.neurokode.com PDO]: Python Database Objects - allows use of most DBAPI modules with a clean, simple API similar in scope to ADO or JDBC. Column access by name is provided by a low overhead mechanism.
PyDo: ORM
[http://www.colorstudy.net/software/webware/SQLBuilder.py SQLBuilder]: SQL generator
[http://dustman.net/andy/python/SQLDict SQLDict]: SQL wrapper
[http://sqlobject.org/ SQLObject]: ORM
[http://roundup.sf.net/ Roundup's hyperdb]: set of constrained data types with relations (many-to-many included) over multiple backends including SQL, metakit and db (yes, relations in an anydbm-backed database
See also PersistenceSystems and ObjectOrientedDatabase