| name | litestar-di |
| description | Auto-activate for Provide, NamedDependency, SkipValidation, Dependency(skip_validation=True), dependencies=, litestar.di, Dishka FromDishka, Inject, or providers. Not for plain parameters. |
Litestar Dependency Injection
Use this skill for Provide, NamedDependency / SkipValidation, dependency
maps, provider factories, request-scoped resources, and Dishka integration.
Code Style Rules
- Mark every name-based injected parameter with
NamedDependency[T]. Use
NamedDependency[SkipValidation[T]] only for trusted provider output that
must bypass validation.
- Wrap providers in
Provide. For synchronous providers, set
sync_to_thread=True for blocking work or False for trivial non-blocking
work; leaving it unspecified emits a warning.
- Use Litestar dependency maps for simple and medium apps.
- Use Dishka when the project needs explicit scopes and provider modules.
- Keep provider names stable and descriptive.
- Do not open request-scoped resources at import time.
Quick Reference
Workflow
- Identify whether the dependency is app, request, transaction, or function scoped.
- Choose built-in dependency maps or the existing DI framework.
- Register providers at the narrowest useful scope.
- Inject dependencies by name or type according to the chosen stack.
Guardrails
- Do not introduce Dishka for one or two simple providers.
- Do not mix dependency naming conventions in one app.
- Do not keep database sessions or clients as global mutable state.
- Do not hide business logic inside providers.
- Do not mutate dependencies on a constructed app in tests. Build a fresh app
or test client with the replacement dependency map.
Validation Checkpoint
Example
from litestar.di import NamedDependency, Provide
async def provide_user_service(db_session: NamedDependency[AsyncSession]) -> UserService:
return UserService(session=db_session)
dependencies = {"users_service": Provide(provide_user_service)}
References Index
Official References
Shared Styleguide Baseline