UI overwhelmed by Websocket messages (?) #77

New Issue

2026-02-15T17:05:58-05:00

yindo commented

2026-02-15 17:05:58 -05:00

Originally created by @padraic00 on GitHub (Apr 24, 2025).

I have an OWUI 0.6.5 helm deployment on an AKS Kubernetes cluster. It's a single replica behind an nginx-ingress (websocket enabled, proxy send/request timeouts set to 10m). Usage is fairly light for now (1-3 users). All models configured on the instance are from external providers (OpenAi, Gemini, etc). Streaming is enabled, but without redis for now.

When I prompt "slower" models, I have an issue where the response stops rendering about 2 minutes into the run. Checking the browser dev console, I can see that, every time this happens, the initial websocket connection closes and another is created. Further, I can see that the client stops receiving ping requests from the server while the response is being streamed, which makes me suspect a ping timeout to be the issue. Could this be the case?

I'll add that this often happens with certain OpenAI models. Checking the websocket messages in one test, I saw that OpenAI was sending messages at 10x the rate of Gemini. Is it possible these messages are overwhelming the stream?

Originally created by @padraic00 on GitHub (Apr 24, 2025). I have an OWUI 0.6.5 helm deployment on an AKS Kubernetes cluster. It's a single replica behind an nginx-ingress (websocket enabled, proxy send/request timeouts set to 10m). Usage is fairly light for now (1-3 users). All models configured on the instance are from external providers (OpenAi, Gemini, etc). Streaming is enabled, but without redis for now. When I prompt "slower" models, I have an issue where the response stops rendering about 2 minutes into the run. Checking the browser dev console, I can see that, every time this happens, the initial websocket connection closes and another is created. Further, I can see that the client stops receiving ping requests from the server while the response is being streamed, which makes me suspect a ping timeout to be the issue. Could this be the case? I'll add that this often happens with certain OpenAI models. Checking the websocket messages in one test, I saw that OpenAI was sending messages at 10x the rate of Gemini. Is it possible these messages are overwhelming the stream?

yindo closed this issue

2026-02-15 17:05:58 -05:00

yindo referenced this issue

2026-02-15 17:15:52 -05:00

[PR #77] [CLOSED] Update index.md for installation method #203

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: open-webui/docs#77