Web Servers in Python

Python web apps talk to servers through one of two interfaces. WSGI is synchronous and serves Django and Flask; ASGI adds async support, WebSockets and long-lived connections for FastAPI, Starlette and Django's async mode. Gunicorn is the long-standing pre-fork WSGI server on Unix: a master process manages several worker processes, and worker classes let it host ASGI apps too. Waitress is a pure-Python, multithreaded WSGI server that also runs on Windows.

For ASGI, Uvicorn is the most common choice and uses uvloop and httptools when they are installed. Hypercorn supports HTTP/2 and HTTP/3 and can run on Trio as well as asyncio. Granian is written in Rust on Hyper and Tokio and can serve ASGI, WSGI and its own RSGI interface from one server. In production, these usually sit behind a reverse proxy or load balancer, with the worker count tuned to the available CPU cores.