The response displayed to the user using the data from the plugin response is sometimes long and drowns the data in its verbosity.
Can the plugin response include instructions//hints to limit the response text length?
It would also be great to force output as bullets or numbered lists.
I have yet to thoroughly test instructions from within the API response, however, placing these instructions in the
description_for_model in the
ai-plugin.json has been working for me.
Both will require some amount of prompt engineering (trial and error).
I’ve had pretty good results with minimal instruction in the
description_for_model when I specifically mention the names of the properties returned by the API response.