Strawchemy

Appearance

Async sessions ​

Strawchemy runs against a synchronous or an asynchronous SQLAlchemy session. This page shows how to switch a schema to async, write async resolvers, and mix both kinds of field.

Async sessions need the asyncio extra: uv add "strawchemy[asyncio]".

Async repository ​

Set repository_type on StrawchemyConfig to make every generated field run async. This option decides whether a schema's fields run through StrawchemySyncRepository or StrawchemyAsyncRepository:

python
from strawchemy import Strawchemy, StrawchemyAsyncRepository, StrawchemyConfig

strawchemy = Strawchemy(
    StrawchemyConfig(
        "sqlite",
        repository_type=StrawchemyAsyncRepository,  
    )
)

Left unset, repository_type defaults to StrawchemySyncRepository, so an async session needs it set explicitly.

WARNING

repository_type and the session it runs against must agree. A schema left at the sync default, run against an async session, builds without complaint — the mismatch surfaces only once a request executes a query, as a GraphQLError in the response:

'coroutine' object has no attribute 'all'

The sync repository passes the session's execute() result straight to code that expects rows back; against an async session that result is an unawaited coroutine. Set repository_type=StrawchemyAsyncRepository on StrawchemyConfig whenever the session is async.

Async resolvers ​

A hand-written resolver picks its own repository class, so it must match the session it runs on. Here is get_post_by_title from custom resolvers, sync and async:

python
from sqlalchemy import select
from strawchemy import StrawchemySyncRepository


@strawberry.type
class Query:
    @strawchemy.field
    def get_post_by_title(self, info: strawberry.Info, title: str) -> PostType | None:
        repo = StrawchemySyncRepository(PostType, info, filter_statement=select(Post).where(Post.title == title))
        return repo.get_one_or_none().graphql_type_or_none()
python
from sqlalchemy import select
from strawchemy import StrawchemyAsyncRepository


@strawberry.type
class Query:
    @strawchemy.field
    async def get_post_by_title(self, info: strawberry.Info, title: str) -> PostType | None:
        repo = StrawchemyAsyncRepository(PostType, info, filter_statement=select(Post).where(Post.title == title))
        return (await repo.get_one_or_none()).graphql_type_or_none()

The async form adds async to the resolver, awaits the repository call, and constructs StrawchemyAsyncRepository in place of StrawchemySyncRepository; the rest of the resolver stays the same.

Mixing sync and async ​

Every field factory — strawchemy.field, strawchemy.create, strawchemy.update, and the rest — accepts repository_type, overriding the schema-wide default for that one field. A schema left at the sync default can still run a single mutation against an async session:

python
from strawchemy import StrawchemyAsyncRepository


@strawberry.type
class Mutation:
    create_post: PostType = strawchemy.create(PostCreateInput, repository_type=StrawchemyAsyncRepository)

Every other field on Query and Mutation still runs through StrawchemySyncRepository; only create_post needs an async session.