I would like to connect an avatar to a locally hosted Large Language Model (LLM), so that you can ask the Avatar questions and it would respond using data from the LLM.

I have a chat/text-based LLM that has privately made parameters that I can ask questions to via chat.  Since you all already have a "Text to Avatar" feature, would you be able to add a feature where a selected Avatar can connect to a locally hosted LLM and can take vocally spoken input/questions from the user and submit it to the LLM and then receive the output from the LLM and have the Avatar speak it to the user?