Description

When using GPT-5 model which currently is a bit sluggish to respond, chances are that one of the requests during response generation times out or is answered with some temporary error. If this happens mid-response, the Agents plugin replaces any already provided output with the following text:

Sorry! An error occurred while accessing the LLM. See server logs for details.

This is bad. Please don't delete already generated responses, just because some subsequent request failed.