What happened?

I am using llama.cpp as backend with infinite timeout (and I also tried with a very high number).
With the command "/settings" I have set http timeout = false into Pi.dev.
But still, i receive a timeout error when using a slow big model.

Is it possible that there is some sort of bug that lets Pi.dev ignoring the client side timeout option? Or maybe there is an additional connection timeout in the http itself?

Screenshot:

Steps to reproduce

1. Load a big model on slow GPU computer using llama.cpp
2. Disable timeout both on llama-server and pi.dev
3. Receive timeouts for every big file read

Expected behavior

No response

Version

No response