Newbie vs Vectorial Database (newbie is losing so far)

Questions
1

Describe the problem/error/question

Hey guys! Im new to coding and i think i bit a little more than I can chew.

One of my clients has a recruitment agency. He asked me to build a worlflow for him that will search for people with similar experience to the open positions he has in his data base (around 3k candidates). These candidates are in a table (ID, Name, PhoneNr, Experience columns).

From what I know, the data should be embedded in vectors and then compared semantically, I’m thinking it will extract candidates that have a certain similarity in their experience compared to the JobTitle.

I have no idea where to start and has not able to find the right resources online so far (probably because I do not know how to search for them). I was hoping someone could help give me a direction about how this flow would look like, what nodes and Agents I should use to be able to identify these candidates?

I looked at RAG agents, but from what I understand it embedds documents as a whole, not each row as in a table. I’m really not sure how to go about it.

Any help is highly appreciated!

PS: I’m using Supabase as a DB and managed to embed the candidate experience in vectors which i put in another column. And I managed to embed the JobTitle. Now i don’t know how to compare them =).

2

Check out the RAG starter template made by the official n8n team. It’s a good start if you are new to vector store.

Just to clarify, vector databases are simply coordinates in the core. Embeddings models compare these coordinates to match close numbers(similar words) and retrieve them as a response. So it doesn’t really matter if it was a table or a document – you need to turn those words into numbers first.

My explanation may lack clarity, so check out the n8n documentation on vector stores for further explanation.

1 Like