There's an interesting stanford webinar covering this topic. It's what they call "Retrieval augmented in-context learning": https://www.youtube.com/watch?v=-lnHHWRCDGk