Chat crashes when submitting a subquery using deepseek-r1:8b when querying a locally served instance of wikipedia

[devyzr]

I have Onyx set up with a local Ollama instance serving Deepseek-r1:8b. It seems to handle a search query OK despite the formatting, but when I click on a generated subquery, it looks like it's processing for a bit, and then the chat closes and a new one is opened.

I'm running Onyx locally via Docker.

You can see the issue happen around the 40 second mark in this video: https://imgur.com/a/4AGaEHi

[devyzr]

This seems to happen when the "Agent" option is selected, and it is still researching all the possible answers. This tells me that it could be an issue of resourcing, where it doesn't expect to have so many document matches.