is there distances_from_embeddings

are you trying to get cosine similarity of 2 embeddings?

yes

So, if you're using a query_engine, the source nodes that come back in the response have a field called score that is the cosine similarity (Default setting) of your query embedding and the node's doc embedding.

Not sure if that fits your use case. Can you elaborate perhaps on what you'd like to do with the such scores?

use if rag fails to get an answer. use a distance approach

Oh maybe I am misunderstanding. Can you kindly elaborate on what you mean by "if rag fails". The RAG system uses cosine similarity by default to find the most relevant documents to the query.

In other words, it uses a distance approach already.