Issue with how AI ‘completed prompt’ is displaying

Hi everyone, first time posting here… hoping someone can help.

Im using the OpenAI integration with the ‘complete a prompt’ function twice, to generate two pieces of text.

The first piece of text is a summary. I then reference this summary in the prompt for the section piece of text (an article).

For some reason (maybe to do with token limits?) the end of the summary (first piece of text) sometimes doesn’t show up on that cell, but appears at the beginning of the article (second piece of text).

Any ideas how to stop this from happening?


Can you show us how you’re constructing the prompts and the OpenAI columns/actions?

What input do you have for the “summary” part? Why do you generate a summary before the article? How are they linked to each other?

Of course.

So I have the summary first, to create a style guide per brand. This is then used in the prompt for a ‘create article’ function.

Prompt to create style guide:

Table structure - includes the prompt column (template) and completed prompt column (triggered on ‘submit’)

This is what a completed prompt cell looks like - cuts off the end of the sentence:

Then I have a separate table to generate each article, which pulls through the style guide from my ‘client’ table as a Relation and Lookup (two columns). Then there’s a prompt (template) for generating the article, and completed prompt column.

In my ‘article’ table, the completed prompt column (where the article appears) starts with the end of the missing sentence:

I’m guessing there’s something about token limits but no idea why it’s showing it like this!