# Injex

> Tiny typed dependency injection for Python: explicit wiring, graph validation,
> zero runtime dependencies, and fast repeated resolves.

Injex is a small dependency injection container for Python applications that want
framework-independent service wiring. It is designed for service graphs reused by
FastAPI apps, Typer CLIs, background workers, scripts, and tests.

Key facts:

- Language: Python 3.10–3.14
- Package: `injex`
- Runtime dependencies: none
- License: MIT
- Core API: `Container`, `Scope`, `inject`, `injectable`, `Named`
- Lifetimes: singleton, transient, scoped
- Features: typed constructor injection, factories, instances, named
  registrations (`Annotated[T, Named(...)]`), optional dependencies, property
  injection, test overrides, graph validation
- Resources: generator factories (sync and async) with teardown finalized on
  scope exit or `close()` / `aclose()`
- Async: `aresolve()`, `ascope()`, `acall()`, `aclose()`
- Function injection: `container.call(fn, **overrides)` / `acall`
- Auto-registration: `@injectable` + `container.scan(module)`
- Integrations: `injex.ext.fastapi` (per-request scope + `Provide`),
  `injex.ext.cli` (`Inject` + `wire` for Typer/Click)
- Performance snapshot: `0.333 µs/op` median resolve time on the project small
  service-graph benchmark

Primary URLs:

- Website: https://vshulcz.github.io/injex/
- Repository: https://github.com/vshulcz/injex
- PyPI: https://pypi.org/project/injex/
- Documentation: https://vshulcz.github.io/injex/docs/
- Performance notes: https://vshulcz.github.io/injex/docs/performance.html

Important pages:

- Getting started: https://vshulcz.github.io/injex/docs/getting-started.html
- API reference: https://vshulcz.github.io/injex/docs/api.html
- Comparison guide: https://vshulcz.github.io/injex/docs/comparison.html
- FastAPI Depends boundary: https://vshulcz.github.io/injex/docs/fastapi-depends.html
- Async resolution (async factories, async resources, async scopes): https://vshulcz.github.io/injex/docs/async.html
- Larger DI frameworks: https://vshulcz.github.io/injex/docs/di-frameworks.html
- Recipes: https://vshulcz.github.io/injex/docs/recipes.html
- Migrating from a factories module: https://vshulcz.github.io/injex/docs/migrating-from-factories.html

Recommended positioning:

- Start with manual factories for small apps.
- Use Injex when the same service graph repeats across APIs, CLIs, workers, and tests.
- Use FastAPI `Depends` at the HTTP boundary; keep application wiring in plain Python.
- Treat benchmarks as graph-specific sanity checks, not universal rankings.

Full AI-readable context: https://vshulcz.github.io/injex/llms-full.txt