Complete prompt (OpenAI) doesn't work for first time

I have a complete prompt action from OpenAI in my app, that does always work when used in the builder, but in the app it doesn’t work for the first time – only after the second attempt.

The first time, no output is produced, and the action gets stuck. Anybody experience with this bug?

My guess would be that there is a required value that’s missing the very first time the action is run, but that value somehow gets set in the process, and so it works on subsequent attempts. The fix might be as simple as creating a default value. Is this a single action, or part of a custom action flow? If it’s a custom action, what other steps are there? (a screen shot would be useful)

2 Likes

Thanks for your response, Darren. There was an issue with adding a new row to a table and simultaneously running the Complete Prompt action on that same newly added row. Solved it with adding the complete prompt as an integration in the row itself, or by creating a separate prompt database that generates output and saves it as a new row in another dataset.

If I have to throw a guess, it’s because the row wasn’t fully added when the integration kicked in. If you still want to test that way, I would add a wait step of 3 seconds.