Paperless-AI output doesnt match direct ChatGPT output

[stratiegery01]

I've been tweaking the prompt settings in Paperless-AI and noticed that no matter what I try, the output of the processed file is most-often incorrect. When I troubleshoot the prompt by directly trying to analyze the same document within ChatGPT, using the same prompt and Open AI model type as in Paperless-AI, the output JSON data is completely different (and GPT produces more accurate data).
So there is a problem with either Paperless-AI not sending the correct JSON data to Paperless-NGX, or the configuration is somehow messed up.

Occasionally when I try to update the prompt itself and try to save the config, I'm presented with this error:

Any thoughts?

[stratiegery01]

This has been resolved. After digging through my docker container logs I noticed that when communicating with chatGPT API, it was getting a token limit error. I paid for more tokens and Paperless-AI began processing documents normally.