I’m trying to execute an action after a timer, from a non-esp32 device.
This script actually works great with chatgpt as llm:
alias: Sarchiapone
fields:
sarchiapone:
selector:
duration:
enable_day: false
name: Sarchiapone
description: sarchiapone
default:
hours: 0
minutes: 1
seconds: 0
comando:
selector:
text: {}
name: comando
description: comando
required: true
sequence:
- service: timer.start
metadata: {}
data:
duration: "{{sarchiapone}}"
target:
entity_id: timer.attesa
- wait_for_trigger:
- platform: state
entity_id:
- timer.attesa
to: idle
- service: conversation.process
metadata: {}
data:
text: "{{comando}}"
description: ""
Funny thing: "sarchiapone¨ is not a real word in italian, based on an old joke about some animal that does not even exist. I HAD to use that nonexisting word.
I can say “sarchiapone for 5 seconds for turning on the light of living room” and it works!
If I change the script name to something that makes sens like “wait”, Assist does not like a sentence like “wait 5 seconds for turning on the light of living room”, it says it can’t start a timer for this device.
This behavior collides with the latest news about supercharging LLMs with our scripts. Theoretically if I create a script called “turn on”, Assist should execute that instead of the basic built-in command. Am I wrong?