AI Endpoint Generates Response, but Nextcloud Never Accepts It

[Poisonsting]

Which version of integration_openai are you using?

v3.7.1

Which version of Nextcloud are you using?

v32.0.1

Which browser are you using? In case you are using the phone App, specify the Android or iOS version and device please.

Firefox 144.0 (aarch64), Safari 26.0.1

Describe the Bug

When using any OpenAI compatible Chat Completion or Text Completion endpoint, this integration connects just fine and lists all available models. It only sometimes sends prompts properly to the backend, but when a response is generated, it never accepts it in the UI.

I've tested using LMStudio and Koboldcpp both as backends. Chatbox AI and Sillytavern both work just fine with either backend.

I'm happy to troubleshoot and submit screenshots as requested.

Expected Behavior

Nextcloud Assistant can be used for AI Chat

To Reproduce

1. Setup an OpenAI compatible endpoint in this integration.
2. Attempt to chat.
3. Receive no response.

[kyteinsky]

hello,
what is your service URL set to and which service are you using?

[Poisonsting]

My service url is set to the local IP and port of Koboldcpp and/or Lmstudio, followed by v1. Something like this: “local-ip:5001/v1”

[kyteinsky]

thanks, that looks good.

It only sometimes sends prompts properly to the backend, but when a response is generated, it never accepts it in the UI.

how do you figure this? Is the CPU or GPU under load after a task is picked?
it would be nice to have some nextcloud server logs. If the response is not accepted, there must be some error logs related to it.

[leo71233826]

ollama with same question, no error but also no reply