Running into some bugs with the GPT creator, here’s what I’ve noticed so far
- Profile Picture
- Cannot upload file (clicking button just closes the dialog)
- Asking ‘Create’ tab to use an uploaded file works briefly, but on reload it gets moved to ‘Knowledge’
- Cannot download a generated profile pic
- Editing / updating an action set is finnicky, if I change the name of an endpoint I typically have to delete and re-create the action set for it to reflect in conversation.
One feature that would be really helpful is the ability to start each chat with a custom system message with content from an API endpoint, e.g. an overview of the user’s profile or documents. Currently I have to provide an action for it and instruct the GPT to call it right away, but this makes starting a chat fairly clunky.