Automation not properly handling openai text response prompt - missing specific instructions

Configuration
, ,
1

Hi community friends,
I’ve taken myself up for a challenge.
I’ve put a door sensor in the mailbox and now would like the sensor to trigger an automation triggering a variable and funny TTS announcement (going through OpenAI) on my home group speakers when the mailman delivers mail.
I’ve been trying with the help of ChatGPT but ain’t getting there. Need some real experts advice here there I’m not very familiar yet with coding and YAML.
Below is my automation YAML-code

alias: Brievenbus Klep Humoristisch Bericht
description: Grappig TTS-bericht bij brievenbusgebeurtenis.
triggers:
  - entity_id: binary_sensor.brievenbus_klep_deur_2
    to: "on"
    trigger: state
actions:
  - data:
      entity_id: input_text.openai_prompt
      value: >-
        De brievenbus klep ging net open. Maak een grappige opmerking over de
        postbode die iets heeft afgeleverd.
    action: input_text.set_value
  - target:
      entity_id: sensor.openai_text_response
    action: homeassistant.update_entity
    data: {}
  - data:
      level: debug
      message: >
        Prompt sent to OpenAI: De brievenbus klep ging net open. Maak een
        grappige opmerking over de postbode die iets heeft afgeleverd.
    action: system_log.write
  - data:
      prompt: >
        De brievenbus klep ging net open. Maak een grappige opmerking over de
        postbode die iets heeft afgeleverd.
    action: rest_command.openai_generate_text
  - delay:
      hours: 0
      minutes: 0
      seconds: 3
  - data:
      level: debug
      message: >
        OpenAI Response: {{ states('sensor.openai_text_response') |
        default('Geen antwoord ontvangen.') }}
    action: system_log.write
  - data:
      entity_id: media_player.home_group
      cache: false
      language: nl
      message: >
        {{ states('sensor.openai_text_response') | default('Er ging iets mis met
        de grap!') }}
    action: tts.google_translate_say
mode: single

Below is the code from configuration.yaml

rest_command:
  openai_generate_text:
    url: "https://api.openai.com/v1/chat/completions"
    method: POST
    headers:
      Authorization: "Bearer [REMOVEDMYKEYBUTISVALIDANDWORKINGBOUGHTCREDITS]"
      Content-Type: "application/json"
    payload: >
      {
        "model": "gpt-4",
        "messages": [
          {
            "role": "system",
            "content": "Je bent een grappige assistent voor slimme huisgebeurtenissen. Je reageert altijd met geestige opmerkingen die specifiek zijn voor de gebeurtenis. Antwoord altijd in het Nederlands."
          },
          {
            "role": "user",
            "content": "{{ prompt | default('Er is geen gebeurtenis meegegeven!') }}"
          }
        ],
        "max_tokens": 50,
        "temperature": 1.0,
        "top_p": 0.9,
        "frequency_penalty": 0.5,
        "presence_penalty": 0.5
      }
    content_type: "application/json"

sensor:
  - platform: rest
    name: OpenAI Text Response
    resource: https://api.openai.com/v1/chat/completions
    method: POST
    headers:
      Authorization: "Bearer [REMOVEDMYKEYBUTISVALIDANDWORKINGBOUGHTCREDITS]"
      Content-Type: "application/json"
    payload: >
      {
        "model": "gpt-4",
        "messages": [
          {
            "role": "system",
            "content": "Je bent een grappige assistent voor slimme huisgebeurtenissen. Je reageert altijd met geestige opmerkingen die specifiek zijn voor de gebeurtenis. Antwoord altijd in het Nederlands."
          },
          {
            "role": "user",
            "content": "{{ states('input_text.openai_prompt') | default('') }}"
          }
        ],
        "max_tokens": 50,
        "temperature": 1.0,
        "top_p": 0.9,
        "frequency_penalty": 0.5,
        "presence_penalty": 0.5
      }
    value_template: "{{ value_json.choices[0].message.content | default('Geen antwoord ontvangen.') }}"
    scan_interval: 3600


tts:
  - platform: google_translate
    language: "nl"

input_text:
  openai_response:
    name: OpenAI Response
    max: 255

So I’m getting back a creative response from OpenAI but it says it’s missing specific instructions (about the mailman delivering mail.)
Respons I’m getting below freely translated:

It looks like you didn't specify a specific event. I'm all set to make you laugh, but I do need a little context. Try

image
image921×293 24.3 KB

When testing, I get these log errors now:

Logger: homeassistant.helpers.service
Source: helpers/service.py:303
First occurred: 5:42:43 PM (2 occurrences)
Last logged: 5:59:00 PM

Referenced entities input_text.openai_prompt are missing or not currently available

and also

Logger: homeassistant.components.rest_command
Source: components/rest_command/__init__.py:154
integration: RESTful Command (documentation, issues)
First occurred: 5:42:48 PM (2 occurrences)
Last logged: 5:59:03 PM

Error. Url: https://api.openai.com/v1/chat/completions. Status code 400. Payload: b'{\n "model": "gpt-4",\n "messages": [\n {\n "role": "system",\n "content": "Je bent een grappige assistent voor slimme huisgebeurtenissen. Je reageert altijd met geestige opmerkingen die specifiek zijn voor de gebeurtenis. Antwoord altijd in het Nederlands."\n },\n {\n "role": "user",\n "content": "De brievenbus klep ging net open. Maak een grappige opmerking over de postbode die iets heeft afgeleverd.\n"\n }\n ],\n "max_tokens": 50,\n "temperature": 1.0,\n "top_p": 0.9,\n "frequency_penalty": 0.5,\n "presence_penalty": 0.5\n}'

Thanks in advance for your (home) assistance :wink: