We use our blog articles as a knowledge base, so I set up a tool in vapi called

get_context

to hit our server and perform the embedding similarity check there, and return the relevant context to vapi.