#Can assist + LLM set an input_text helper ?
1 messages · Page 1 of 1 (latest)
I have had great success with LLM using scripts to fulfill tasks. Just create a script with an input field, a descriptive name ("Play music") and an even more descriptive description ("this script will play music. provide the title in the input field 'title'"). within the script fill the input_text. then expose the script to Assist and test it.
I've created multiple scripts (with inputs if necessary) that I can trigger via Assist without exposing all the details of the entities involved in the script. great from a privacy perspective but also great because YOU define how something is done and merely let the LLM figure out THAT and HOW it can use this script but nothing more. very predictable results!
"Descriptive" is the key word above
Works flawlessly. Will replace the input text with my actual backend logic. Thank you!