OpenAI + Zapier Unpredictable Results

Hello Fellow Astronauts (we’re pushing boundaries, here!)

Straight to the questions:

*Does anyone know of a way to force completion of a Zap step? *
*Verify completion? *
*Re-attempt if fail? *
Anything that can increase my reliability?

For my project needs, its best if I have an ongoing conversation with an Assistant. I’ve created one on OpenAI but Zapier cannot seem to find it, I think I’ve tried everything including passing the link directly to the Zap for them to locate the Assistant. Since those efforts have failed, one of my first steps in the Zap is to create a new Assistant, I pass it all the same instructions that my OpenAI-hosted Assistant contains.

So I guess my first question is how do I manage token usage, or make sure that my Zapier-hosted Assistant (the one created after the Zap is triggered) doesn’t get shut down or limited for over usage? I’m collecting prompt and completion tokens at the end of my process and I’m easily hitting 300k tokens (total) on practice runs.

Back to the beginning, after Zapier creates a new assistant, the zap engages a continuous conversation with that assistant. But it seems like the Zap or OpenAI is either timing out or otherwise generating errors. Sometimes my output is perfect, sometimes there are obvious failures. I haven’t changed anything.

Hey there!

So, I’ll be the first one to admit that it’s tough to tell whether or not this is an OpenAI API issue or a Zapier issue.

If it were me, I would go and look through Zapier’s own development community stuff and see if there’s anything similar (or somewhere to post). I would rule out it isn’t anything from Zapier first, then I’d come back here.

You can find their dev community platform here: https://community.zapier.com/p/developer-zone