Workarround for prompt limits to add good and bad examples

[wisefinger]

Hi,

I have added both good and bad examples to a prompt description and this works fine. However, I was wondering how I could use learnings of examples that where evaluated in an annotation queue. I guess I could copy them into the prompt as test but these is a limit for the tokens. Does anyone used different ways? I was thinking about adding a tool to the agent that allows to search the examples for similar cases although I'm not sure how well this would work. Ay thoughts or ideas?

Kind regards,
Kris