[Bug]: Ollama integration behaving strangely

[FamousL]

What did you do when it broke?

1. Configure ollama, used llama3.2 for the general conversation ai.
2. Create notebook
3. Add content
4. Attempt to talk to AI inside the notebook

How did it break?

Ai window pops notification - "Failed to send message", wait 5 minutes reload, message is in the window, along with the ai's response

I expected no such error, though that may require a longer timeout.

Logs or Screenshots

No response

Open Notebook Version

podman version 5.6.2

Environment

Fedora 43 ollama version is 0.11.0 podman version 5.6.2

Additional Context

No response

Contribution

• [ ] I am a developer and would like to work on fixing this issue (pending maintainer approval)

[lfnovo]

Can you confirm this happens even after you load the model to memory? Like second, third messages too?

[FamousL]

I can confirm that, it happens with every message. I asked ollama to keep the model in memory as well, but that does not seem to have made a difference.

[lfnovo]

Thanks. Can you also provide some logs for the container? Perhaps there is something interesting to help point to the root of the issue.

[FamousL]

Edit: removed failed upload, next comment has the correct one.

Apologies for the delay, here's a log where it at first didn't happen, but a subsequent query has failed. I did not check this at all, I do not know if there is anything in here that I need to change by sharing this.

[FamousL]

logs.txt

sorry that doesn't look like it loaded properly, trying again.

[hackdefendr]

If it helps, I can add to this as confirmed. I see this as well with Ollama running on my Raspberry Pi 5 but with the Qwen3 model. Specifically on the Insights generation page after adding a simple URL for a source, it just pops up saying Failed.

[abstractprophet]

I can confirm as well. I get the failed to send message error when asking questions to the notebook chat. I have ollama running in python while open notebook is in desktop Docker on a Windows rig.

[FamousL]

I think it is worth noting that after giving ollama some time, and reloading the page does show a response, the chat updates with both the question asked, and an answer.