m0serve Realtime from a synchronous Python app

m0serve

Realtime from a synchronous Python app, with no added infrastructure.

A plain sync Django or Flask view can hold a Server-Sent Events stream or a WebSocket by answering with two response headers, and reach every subscriber on every worker with one function call. No Channels, no Redis, no daphne, no second process. The server is written in Mojo and installs as one wheel with no dependencies.

from m0serve import m0pub


def events(request):                    # an ordinary synchronous view
    r = HttpResponse(": connected\n\n", content_type="text/event-stream")
    r["M0-Hold"] = "stream"             # m0serve holds the connection open
    r["M0-Channel"] = "news"            # ...subscribed to this channel
    return r                            # Django's part in it ends here


def announce(request):                  # any view, command, or cron job
    m0pub.publish("news", "deploy finished")
    return HttpResponse("ok")

pip install m0serve
m0serve myproject.wsgi --realtime

See it: demo.m0serve.dev. Open it in two tabs and type in either; the other shows it within the second, over a held SSE stream and over a WebSocket, from one file of sync Django with channels namespaced per visitor. The line at the bottom says which m0serve version is serving it.

The name is spelled with a zero, m0serve, the way the packages underneath are named: m0-core, m0-http, m0-wsgi.

The view runs first, with sessions and permissions in hand, which is where your authorization belongs. Under gunicorn the two headers mean nothing (they pass through to the client) and the same view degrades to a short plain response, so adopting it does not fork your codebase.

The Flask version is the same four views plus one flag on the socket route, websocket=True, which Werkzeug's router requires before it will match an upgrade request; the Quickstart's §7 is that file, and CI runs it.

Where to go

  • Quickstart: ten minutes from pip install to live multi-tab sync from one Django file. CI executes every command in it on every pull request.
  • Running m0serve: which flags for which app, the execution modes, what to put in front of it, how it shuts down.
  • Capabilities: what the server does, one row per capability, each naming the test that proves it.
  • All documentation: the map, by what you are trying to do.

What it also does

  • Serves WSGI and ASGI from one binary. The protocol is detected from the object. Django, Flask and FastHTML each have a smoke test in CI, and the quickstart's Flask realtime views are driven by the same RFC 6455 probe as Django's; PEP 3333 conformance is validated by wsgiref and ASGI by a validator written from the spec.
  • Several applications in one process, each in its native mode: m0serve --mount /=shop.wsgi --mount /app=live.asgi runs sync Django on handler threads and async FastHTML on an asyncio executor behind one listener with one shutdown.
  • A slow view does not stall the rest. Handler threads behind each event loop are the default for WSGI, so a 2 s view no longer holds the keep-alive connections pinned behind it.

What it is not

  • No TLS and no HTTP/2. Terminate at a proxy, as with gunicorn.
  • Not the fastest server on raw throughput. The benchmarks say where it loses and by how much, and every figure there is recomputed from a dated artifact. What it wins is the fast-request tail under mixed load.
  • Pre-1.0. The API can still change; the changelog records every change.
  • macOS arm64 and Linux x86_64 and aarch64 only. No Intel Mac, no Windows, no musl. CPython 3.10 to 3.14, free-threaded builds included for WSGI.

For agents

llms.txt is the operating contract in one page, with every page of this site indexed as Markdown; /llms-full.txt is the essential pages in one file; spec.json is the capability matrix as data.

Underneath

m0serve is one package of mojo-http, an HTTP/1.1 server and web framework for Mojo with routing, content negotiation, ETags, SSE and WebSockets, a Datastar adapter and SQLite bindings. The README covers that side; the source is on GitHub under the MIT license.