Chat Prompt Evaluation Unable to Reference Prompt ID or Saved Prompt on platform

API Bugs
1

When creating a prompt version in chat prompt section on OpenAI platform we can navigate to Evals through the button on top right - set graders and custom dataset. I am trying out this feature to see if we can update prompt versions and also test the performance on the dataset right after on the platform. The issue I am running into is that when I navigate from the prompt version to Evaluate, attach a dataset loaded in the Storage, the platform is unable to supply the prompt template that I came from thus forcing me to reenter the whole configuration.

Similar issue comes when dealing with evals API where I would like to be able to reference a prompt by prompt id which seems to not be currently supported. Is there any development on this?

Thanks a lot for a response.

2

Moreover, when trying to import logs for evaluation I would really appreciate filter by prompt id that was used to generate that log. For now I would have to bypass the missing filter by passing prompt id to request metadata.

Screenshot 2025-09-24 at 14.54.25
Screenshot 2025-09-24 at 14.54.25537×896 11.2 KB

3

I’m having this same issue. Oddly enough, I have a prior example eval that I was able to create through the Evaluate button (via Chat –> Create). I can see some metadata was added to the eval for the specific prompt it relates to.

I feel like this is meant to be used in an AI ops flow, where we’re able to create an Eval Run against a version of a model before we roll it out.