LiteLLM caching interferes with "refreshing" responses from GPT, Claude, etc.

[rcdailey]

I use LiteLLM behind Open WebUI. In the UI, when I click the "refresh" button on an AI response, it's supposed to resubmit the prompt but provide a different randomized output. I usually use this when I'm not happy with the response I got. However, I guess due to the fact that I've enabled the Redis caching, it instantly responds with the exact same text.

Is there a way to configure caching to be a little smarter about intentional forced resubmits? Thanks in advance.