Veloce

Async Python web framework where one route definition drives an HTTP endpoint, an OpenAPI operation and an MCP tool — ASGI-native, batteries included.

Documentation: veloceframework.com

Source code: https://github.com/Lokesh-Tallapaneni/veloce

Veloce framework (PyPI: veloceframework) is an ultra-fast async Python web framework — ASGI-native and batteries-included.

Veloce is a from-scratch async Python web framework. The router,
request/response pipeline, dependency injection system, OpenAPI
generator, WebSocket layer, and test client are all in-tree — not
wrappers around an existing stack.

Position

Veloce is a from-scratch async Python web framework — not a wrapper around Starlette, FastAPI, or Flask. It provides FastAPI-style typed dependency injection, Pydantic v2 request/response validation, OpenAPI 3.1 schema generation, WebSocket support, and Flask-compatible helpers (g, flash, blueprints, session) in a single tree. It is built on a radix-tree router, an in-memory test client, and built-in CORS/CSRF/session/rate-limit/security-headers middleware. An external security review of the request path is planned ahead of 1.0; until then, treat it as pre-production software.

Key design points

• Async-first handlers. async def handlers run directly on the
  event loop; sync def handlers are supported transparently and run in
  a thread-pool executor so they do not block it.
• Precompiled dispatch. Handler signatures are inspected once at
  registration into a HandlerPlan; the per-request hot path performs
  no reflection.
• Radix-tree routing with typed path converters (int, float,
  string, uuid, path, date, datetime, time, timedelta,
  decimal, any, plus custom).
• Typed dependency injection via Depends, Security,
  SecurityScopes, including yield-style dependencies with teardown.
• OpenAPI 3.1 generated from Pydantic models, served through
  Swagger UI and ReDoc out of the box.
• In-memory test client drives the real ASGI surface — no sockets,
  no separate server process.

Requirements

Python 3.10+.

Veloce depends on orjson, httptools, pydantic v2, python-multipart,
multidict, jinja2, and uvloop on non-Windows platforms.

Installation

pip install veloceframework

That alone serves, via the built-in app.run() server. For a deployment, the
standard extra adds uvicorn, brotli and msgspec:

pip install "veloceframework[standard]"

The installed package exposes the veloce import:

from veloce import Veloce

Example

Create main.py:

from veloce import Veloce

app = Veloce()


@app.get("/")
async def index() -> dict[str, str]:
    return {"hello": "world"}


@app.get("/items/{item_id:int}")
async def read_item(item_id: int) -> dict[str, int]:
    return {"item_id": item_id}

Run it:

veloce run main:app

veloce run serves under uvicorn when it is installed and falls back to the
built-in server, which needs no extra dependencies. To pin the built-in server
regardless, call app.run().

uvicorn is an optional extra — install it (pip install veloceframework[uvicorn])
to serve under it or any other ASGI server, or use app.run() / the gunicorn
VeloceWorker:

uvicorn main:app

Open http://127.0.0.1:8000/items/42 to see {"item_id": 42}, or
http://127.0.0.1:8000/docs for the interactive OpenAPI UI.

Feature surface

The full Tier 0/1/2 feature matrix and per-feature design notes live in
docs/.

Project status

Veloce follows semantic versioning: the public
API surface — the names exported from veloce/__init__.py — is what each
release commits to. The current version is shown by the PyPI badge above.

Sponsor

Veloce is developed in the open. If it is useful to you, you can support its
continued development through
GitHub Sponsors.

License

MIT. See LICENSE.