Help needed with embedding recommendations and multiple databases

General API discussion
1

Hey all!

I’m looking to build a tool with the Embeddings endpoint, specifically recommendations. But I’m not sure how to setup the database that I want to use for the recommendations.

My database consists out of connections that each have:

  • Name
  • Location
  • Description
    (But here comes my question) they also have
  • Job expetiences
    That in return have:
    — period
    — job title
    — company names
    ------- company descriptions
    ------- location

How do you use these type of database structures and building things with the embedding endpoints.

Help needed and much appreciated!!

2

do you found a solution for your question? im have the same one!

3

From what I understand embeddings simply calculate and give similarity score

I don’t know if this is right, but I don’t think you can just feed it database tables that have no context

You and the devs know about the context, but any outsider would just see a bunch of rows with random data

I think you’d need to generate contextual output, and then feed that to the embeddings tools

… curious if there is another way to do this (IE. simply describe the rows, columns, and let the ai figure it out) :thinking: