error 400 at small changes

[Sethakan]

Hello. Im trying to build ai automation for my homebattery an ev. I do this with my homey pro 2023 and chat gpt. I have som problems. The the following text is ai generated. I know nothing about coding.

Here is a clear, concise GitHub-ready issue description you can copy/paste:

---

Title:

OpenAI “Ask ChatGPT… and WAIT” card clears prompt & sends empty messages → 400 error

---

Description:

I’m experiencing a consistent issue with the “Ask ChatGPT the question … and WAIT for the full answer” card in Homey.

Whenever I edit the prompt text in the card (even small changes), the card often breaks and the OpenAI API returns:

Error 400: Invalid 'messages': empty array. Expected array with minimum length 1, but got an empty array instead.

This means the card is sending:

{
  "messages": []
}

instead of the original long prompt.

What seems to happen:

• The card uses one single text field for the entire prompt.
• When the text becomes long (multi-section system rules, JSON-templates, etc.), saving or editing the field frequently causes Homey to silently clear the internal messages list before the request is sent.
• As soon as Homey sends messages: [], OpenAI returns 400 and the flow becomes unusable.

Reproduction:

1. Create a flow with the “Ask ChatGPT the question… and WAIT” card.
2. Paste a long prompt (multiple sections, rules, JSON instructions).
3. Save → works initially.
4. Reopen the card and add even a single new line or small change.
5. Save again → next execution throws:
   Error 400 Invalid 'messages': empty array.

Expected behavior:

• The card should preserve the internal message array even when editing long prompt text.
• Saving the card should not produce an invalid empty messages array.

Actual behavior:

• After editing the text field, the card sometimes resets the internal messages array to empty, causing all API calls to fail.

Impact:

• It becomes impossible to maintain or update complex prompts (e.g., Home-energy optimization logic, multi-rule system prompts, structured JSON outputs).
• The card becomes unusable for advanced automation.
• The only workaround is to completely re-paste the entire prompt every time and hope the save doesn’t clear the messages array.

Request:

Please investigate why editing the text field causes the card to send an empty messages array. Ideally, the card should always:

• treat the text field as the system message, or
• expose explicit “system” and “user” message fields (similar to OpenAI Pro),
• and never send messages: [] under any circumstances.

---

If you want, I can also help you write a short repro flow or attach simplified test prompts.