How to emit(through call method) socket events from Celery tasks to clients connected via socketio.ASGIApp

[Trinkes]

Hello everyone

I'm working with a FastAPI backend using socketio.ASGIApp to serve WebSocket connections. I also have a Celery worker running background tasks. I'd like those background tasks to emit socket events to clients connected to the FastAPI server.

Setup:

• FastAPI + socketio.ASGIApp
• Celery worker (running separately)
• Redis used as the message queue

Goal:
Emit an event from a Celery task to a client currently connected to the FastAPI + Socket.IO server, and wait for an ACK, which is why I’m using the call() method instead of emit().

What I’ve already tried / read:
I’ve checked the following:

• GitHub Discussion #809
• Message Queue documentation

While those approaches work for emit(), they don’t support call() — since AsyncRedisManager and RedisManager don’t provide the call() interface needed to wait for a response from the client.

Question:
What’s the recommended way of sending socket events (with ACKs) from a Celery task in this setup?

Should I:

• Create a new Socket.IO server and client for communication between Celery and the FastAPI server? (Feels like overengineering.)
• Use Redis directly to emit the event, with both FastAPI and Celery sharing the same Redis instance and message queue?
• Something else?

If there’s a recommended pattern or best practice for this case, I’d love to hear it — and code snippets would be very helpful too.

Thanks a lot!

[miguelgrinberg]

The limitation here is that your Celery process is not a server process, so it is not set up to listen, which is necessary to receive the callback notifications that the call() method uses. Without the ability to listen, there is no way to do this.

The intention of this feature is to provide a quick way for a non-server process to send a message to a client through the server. If you need more than that, then you will have to use a different solution.

For example, your Celery job can connect to the FastAPI server as a Socket.IO client (using the Client or AsyncClient classes in the python-socketio package). Then it can talk to the server directly via WebSocket, which can listen for messages or callbacks.

[Trinkes]

thank you for the reply @miguelgrinberg .
I wanted to avoid that approach since it's one more connection to deal with but I think it's the simpler approach, I'll move forward with it.