Another question: How detailed should the context commands be? should we describe the entire scene to Neuro with them, or will she have some access to her AI vision and we can somewhat rely on that?
She won’t be using vision most likely.
