ORM in Python

SQLAlchemy is the most capable Python ORM and has two layers: Core, a SQL expression language, and the ORM, with a unit of work and identity map on top. It supports many databases, has an asyncio extension, and pairs with Alembic for migrations. The Django ORM is tied to Django but is the simplest path inside it, with models, querysets and migrations built in. Peewee is a small, expressive ORM for projects that want less machinery. SQLModel, by the author of FastAPI, combines SQLAlchemy and Pydantic so that one class can be both a table model and an API schema.

For document and key-value stores, MongoEngine is a synchronous object-document mapper for MongoDB, while Beanie is async and built on Pydantic models, which suits FastAPI projects. django-mongodb-backend is MongoDB's official backend for using MongoDB through the Django ORM, and PynamoDB gives Amazon DynamoDB tables a model-based Python interface. Document mappers add structure to schemaless stores, but they cannot make a document database behave like a relational one, so model your data around the queries you need rather than translating relational tables one to one.