@autumn i dont know why home assistant isnt exposing more data to the llm like i have a weather entity exposed to it but its only able to get temp and humid even though the entity provides a daily and hourly forecast
@karebu idk if this is a bug or if it just Doesn't Know what to do, idk how long they've been developing the assistant stuff for it but it would probably be pretty difficult to do.
@autumn from what ive seen online for some things you have to create a script/automation that it can call because apparently there just isnt many intents available to the llm, i think custom intents can be written in the config but idk
@karebu thats probably something ill look into when i get my own place, right now i barely have any smart devices and just wanted the voice preview because its smth cool to play around with + wanna use it in my future setup
@karebu yeah im having trouble with that, sometimes itll spit yaml at me too. i tried to ask it a question about home assistant OS and i think it was giving me json about the supervisor or smth?? it was so weird.
@autumn i think this just depends on the quality of the model youre using , ive had zero issues with gemini flash 2.0 besides home assistant just not letting it do enough
@karebu im using llama3.1 8b which is what they said they used in testing, but i should try other models too. the limitation is defo HA but ig its still getting developed etc.
i found the speech to text isnt perfect either but sometimes the LLM will be able to make sense of it even if the correct words weren't used.
