Autumn

A modern ASGI web framework focused on typed controllers, dependency injection, and clean configuration.

Autumn is a Python web framework for building HTTP APIs and WebSocket applications with a small, explicit core. It leans on Python's type system instead of large decorator stacks: route parameters are typed, request bodies are inferred from Pydantic models, dependencies are injected from signatures, and responses can be serialized automatically.

If you want class-based controllers, typed configuration, built-in dependency injection, OpenAPI/Dependencies docs generation, and CORS support, Autumn is built for that style.

Highlights

• ASGI-first application object that works with standard ASGI servers such as uvicorn
• Class-based REST controllers with typed path parameters like {id:int} and {file:path}
• Signature-driven dependency injection with @service and @leaf
• Explicit module discovery without recursive filesystem execution
• Automatic request body validation from Pydantic annotations
• Automatic JSON serialization for Pydantic return values
• Built-in configuration system with environment, JSON, and YAML sources
• Built-in configs for application settings, CORS, and WebSocket tuning
• OpenAPI and dependencies documentation generation with built-in viewer
• Middleware hooks, lifespan hooks, file responses, redirects, and streaming
• WebSocket routes with dependency injection support

Why Autumn

Autumn tries to keep the ergonomic parts of modern Python frameworks while staying direct:

• Controllers are just Python classes.
• Dependencies come from constructor or method signatures.
• Request bodies are inferred from type annotations instead of extra decorators.
• Configs are plain Python classes with typed fields.
• The framework stays close to raw ASGI concepts when you need to drop lower.

That makes the happy path concise, while still keeping the codebase readable when the application grows.

Benchmarks

Latest repeated local benchmark run: 2026-08-12

Environment:

• Windows
• Python 3.12
• uvicorn --workers 1 --loop asyncio --http httptools --lifespan off
• Concurrency: 64
• Warmup: 2s per scenario
• Measurement duration: 8s per scenario
• Repetitions: 5, randomized framework and scenario order
• Reported values: median of repetitions

Median RPS by scenario:

Framework	Plaintext	JSON	Path parameter	Validated body
Falcon	8381.56	8347.52	8391.16	6967.47
Starlette	8403.90	8333.73	8264.93	6748.19
Autumn	8107.42	8223.67	7993.92	6866.13
FastAPI	7470.63	6787.41	5810.57	5277.82

In this run, Autumn is 23.06% faster than FastAPI on average, 1.76% behind Starlette, and 2.79% behind Falcon. The comparison contains 80 measurements in total; every measured request completed successfully.

Explicit module discovery

Autumn never scans and executes every Python file below the project root. Modules that contain independently decorated controllers, services, lifecycle hooks, or configuration are listed explicitly:

app = Autumn(
    discover = (
        'project.controllers',
        'project.services',
    )
)

When root_path is provided, module names are resolved relative to that directory. When a discovered name points to a package, Autumn loads every Python module and nested package inside it. This makes discover = 'project.controllers' sufficient for a controllers package with multiple files. Discovery stays within each explicitly listed package; imports made by its modules continue to work normally.

Philosophy

Autumn favors:

• strong typing over implicit magic
• signatures over decorator-heavy ceremony
• built-in primitives over mandatory third-party integration
• readable application structure over framework cleverness

The goal is to make small apps pleasant and larger apps maintainable.

Author

     _      _  _               _ _   
  __| | ___| || |   ___  _   _| | |_ 
 / _` |/ _ \ || |_ / _ \| | | | | __|
| (_| |  __/__   _| (_) | |_| | | |_ 
 \__,_|\___|  |_|  \___/ \__,_|_|\__|

Thank you a lot!

How to reach me