Hi,
I just thought about making the voice iteration with home assistant more smarter. By adding an additional conversation agent in the pipe line could do it. The first conversation agent would be the OpenAI with option to format the message. The text message comes from user, would be formated with a predefined instructions for OpenAI to allow the response be feeded into the home assistant conversation agent. Some responces would pass as ignored tag to the output text for delivering the OpenAI answer to the user. For more advanced usage would be two tags messages from OpenAI ( first conversation agent) one for success of the feeded command and second for failure. Sorry if this sounds complicated. 
1 Like
I am on the same page as you and trying to figure this out…
I’m also interested. Any progress on this?