#Creating a "home summary" with Claude

1 messages · Page 1 of 1 (latest)

weary sparrow
#

I'm super enthousastic about the new dashboard titles/headers, and I thought it would be a great place for an AI summary about the state of my home. But below automation only works half the time, and I cannot find out why. I get a result from Claude, but I cannot seem to set it in the input_text. What am I doing wrong?
I added the rule out any time-out issues, I'm pretty sure it's not necessary.

id: '1741211701854'
alias: Set text
description: ''
triggers:
  - trigger: time_pattern
    hours: '*'
conditions: []
actions:
  - action: conversation.process
    metadata: {}
    data:
      text: "<<< PROMPT >>>"
      agent_id: conversation.claude
    response_variable: response
  - delay:
      hours: 0
      minutes: 0
      seconds: 30
      milliseconds: 0
  - action: input_text.set_value
    metadata: {}
    data:
      value: '{{response.response.speech.plain.speech}}'
    target:
      entity_id: input_text.summary
mode: single
heady badger
#

The response might be too long to fit in the input helper.

#

Try to cap the response below 255 characters: …plain.speech[0,254]}}

weary sparrow
#

thanks, will try that. i did not realize the input_text had this limitation

weary sparrow
#

Apparently it is quite hard to limit the response of an LLM to 255 characters. What would be an alternative way to store the response before showing it on my dashboard?

orchid flame
#

Have you tried to tell the LLM in your prompt that for API reasons, it's response must be no more than 255 characters, including spaces and punctuation, so it needs to be succinct and not simply truncate the response?

slate pine