Troubleshoot ChatGPT in AutoResponder
Check out the general ChatGPT instructions here.
You keep getting a similar response even though you have changed the prompt? GPT also takes into account the reply history if used in the prompt. To fix the error, you can, for example, delete the reply history for this AutoResponder rule.
Your replies are cut off at a certain length? Increase the max_tokens parameter or tell the AI to write shorter answers.
If you receive the following error, this can have several causes.
Error: One of the parameters is invalid, your OpenAI account needs your attention or OpenAI is currently unavailable.
Please test the ChatGPT feature first with the default settings as described in this guide and as they are pre-filled by AutoResponder. If it still doesn't work, please don't give up.
Check ALL these steps please:
Make sure your API key is valid.
Make sure that a credit card is connected here (You'll only be billed at the end of each calendar month for usage during that month).
Make sure you didn't hit your maximum monthly spend (hard limit), otherwise increase it.
Your OpenAI account needs to be a member of an organization to use the API.
All parameters need to be correct as described in the GPT documentation (e.g. the model name, temperature setting etc.).
Make sure the prompt+response combined doesn't exceed the supported 4096 token limit.
Check your internet connection.
When using a different model, please check if you have already access to that model. Some require you to sign up for a waitlist on OpenAI. GPT-4 is currently only accessible to those who have made at least one successful payment through the OpenAI developer platform.
In case you face a problem when using a fine-tuned model, please contact me via mail.
Pay special attention to connecting a credit card here if you don't have any free OpenAI credits left.