Container

The container is the object responsible for creating and managing your application's dependencies. When you request a dependency, whether via type hints, container.get(), or a framework integration, the container builds it and any dependencies it requires.

In practice, you mainly interact with the container directly during setup. Once configured, dependencies flow automatically through type hints and decorators.

Creation

Wireup provides two ways to create containers, depending on whether your application is synchronous or asynchronous.

create_sync_container

Use this for traditional, blocking Python applications (e.g., Flask, Django, scripts).

container = wireup.create_sync_container(injectables=[...], config={...})

create_async_container

Use this for async/await based applications (e.g., FastAPI, Starlette). It supports async factories and has async methods for retrieval and cleanup.

container = wireup.create_async_container(injectables=[...], config={...})

Arguments

Both creation functions accept the following arguments:

Argument	Type	Description
injectables	list[Union[type, Callable, ModuleType]]	Classes, functions decorated with @injectable, or modules to scan. Modules are scanned recursively, collecting only items decorated with @injectable. Recommended default for production apps: module/package scanning. For environment/flag-based assembly of injectables, see Conditional Registration. For reusable provider-style wiring (same graph, different runtime settings), see Factories: Reusable Factory Bundles.
config	dict[str, Any]	Configuration dictionary. Values from this dictionary can be injected using Inject(config="key").
concurrent_scoped_access	bool	Set to True if you share scopes across multiple threads/tasks. Defaults to False. See Lifetimes & Scopes: Concurrent Access for details.

Multiple Containers

The @injectable decorator only stores metadata, it doesn't register anything globally. Each container you create is fully independent with its own state.

Core API

get

Retrieve an instance of a registered injectable.

SynchronousAsync

db = container.get(Database)
readonly_db = container.get(Database, qualifier="readonly")

db = await container.get(Database)
readonly_db = await container.get(Database, qualifier="readonly")

Qualifiers: Use the qualifier argument to retrieve specific implementations when multiple are registered. See Interfaces for more details.

Important

Prefer constructor-based dependency injection over calling get directly. Use get as an escape hatch for advanced scenarios like dynamic service lookup or when working with framework integration code.

close

Clean up the container and release resources. This triggers the cleanup phase of any generator-based factories.

SynchronousAsync

container.close()

await container.close()

enter_scope

Create a scoped container. Scoped containers manage their own scoped and transient dependencies while sharing singletons with the root container. See Lifetimes & Scopes for details on how scopes work.

You can optionally pass a positional mapping of pre-created instances to provide at scope entry:

with container.enter_scope({DbSession: existing_session}) as scoped:
    db_session = scoped.get(DbSession)  # Uses provided instance

See Advanced: Provided Instances for caveats and ownership rules.

SynchronousAsync

with container.enter_scope() as scoped:
    db_session = scoped.get(DbSession)  # Fresh instance per scope

async with container.enter_scope() as scoped:
    db_session = await scoped.get(DbSession)  # Fresh instance per scope

See Lifetimes & Scopes for details.

config

Access configuration values directly from the container. This provides programmatic access to the configuration dictionary passed during container creation.

env = container.config.get("app_env")
db_url = container.config.get("database_url")

Important

Prefer Inject(config="key") in dependency constructors over accessing container.config directly.

override

Substitute dependencies for testing. Access via container.override.

with container.override.injectable(target=Database, new=mock_db):
    ...  # All injections of Database use mock_db

See Testing for details.

Eager Initialization

By default, objects are created lazily when first requested. To avoid latency on first request for expensive services, initialize them at startup:

for dependency in [HeavyComputeService, MLModelService, Database]:
    container.get(dependency)  # or `await container.get(dependency)` for async

Next Steps

• Lifetimes & Scopes - Control how long objects live.
• Factories - Create complex dependencies and third-party objects.
• Testing - Override dependencies and test with the container.