#Questions about general state of things & Agent functionality

1 messages · Page 1 of 1 (latest)

compact relic
#

Hey everyone. I'm currently abroad so I can't play with things and see for myself, but I'd love to hear your thoughts and experiences on local voice assistants with LLMs. I have a 3080, so I'm pretty limited to the smaller end of models. What are the capabilities? I understand that there is some smart home controls are possible, but how scripted are they? Do we have to define everything ourselves and the voice assistant just matches those sentences with controls, or is there some google assistant / gemini-like decision making by the LLM to figure out what the sentence means (which room, which devices, what controll, etc.) before deciding on an action? Also, i really like using LLMs to ask them about stuff and get relatively recent information, i.e. web search functionality on the side of the LLM. I've seen some stuff about using agent interfaces (?) in between HA and the LLM so that they can decide to do some actions before replying back, but how well does that work?