I have an issue with Hume right now. if you interrupt the agent, it resets itself, completely forgets what it was saying and then starts a new sentence related to what you said. the problem is that if you speak back to the agent saying something like what? the the agent resets again and you're going around in circles with the agent.

I believe its an issue with the user_interruption firing as soon as the user says something like simply agreeing with the ai agent.

normal conversation doesn't flow this way. if you interrupt someone in conversation, normally the person that is interrupted finishes their thought, listens to what they said and then finally responds.

this is what I need:

However, the core behavior you're describing—EVI stopping and responding to interruptions—is by design. (1) If you need EVI to "finish its thought" before responding, that would require a different interaction model than the current interruptibility feature provides.