Use "private" dataset as basis for AI responses

It occurred to me that if you have a library of 5000 examples you can use semantic search to pick 2 or 3 that are close to the new one you want to generate to use as samples.

1 Like