Interactions via buttons like summarize, suggest questions, the questions itself should be exempt from the rate limit as it’s really awful UX to switch to Llama when that button is already here and tempting to just click and move forward. And as the result you’re cut off when not expecting. During the text input it’s much more predictable and feels better; also switching before chatting isn’t that big deal.
PS Also the bound to a centralized solution for paying for Brave leaves you questioning the ethos.