Bug Report: ChatGPT Plugin "Halt/Stop" Feature Not Working After API Requests

Description: After making an API request with the ChatGPT plugin, the ability to halt or stop the chat’s progress is unavailable. Instead, the UI that is typically displayed upon chat completion is shown.

Impact: This issue poses a significant problem, leading to unnecessary delays when a chat needs to be stopped for any reason. The problem is particularly troublesome during the development phase when quick iteration is crucial. This bug’s severity is rather high since we are limited to one chat at a time. However, as there is a workaround (detailed below), it may not be considered an immediate priority.

Steps to reproduce:

  1. Initiate a chat where the ChatGPT plugin makes an API request.
  2. Observe that the “Halt/Stop” feature is unavailable after an API request even though ChatGPT is still generating text or additional API requests.

Expected behavior: The “Halt/Stop” feature should remain accessible after the first API request, allowing users to stop the chat’s progress as needed.

Current workaround: Click the “New Chat” button and wait a few seconds. This seems to halt the current chat effectively.

Thank you!

4 Likes

Flagged to the team, thank you! We have lots of high priority things going on but will hopefully tackle this soon.

3 Likes

Just tried to re-produce this along with @andrey-openai and it seems to be working as expected. Please let us know (maybe with a video) if you continue to have this issue.

@logankilpatrick, this is still not working as expected on my end.

This tweet shows the issue: https://twitter.com/wfhbrian/status/1654515765633077249

The same thing is still happening.

@logankilpatrick This bug has become more problematic since plugin usage counts as additional messages toward the usage cap.

The ChatGPT UX is the same as in the tweet I shared prior, except that the “Regenerate response” button is no longer present. The screenshot below, taken mid-request creation, shows this.