Looking for examples of the native minhash being used for near duplicate detection

Following the documentation examples, I have mapped a text field for minhashing of shingles. Using it in a more-like-this query does work, but, as the documentation says, I'd like to use it for Jaccard similarity and retrieve that value. Is it possible, and if so, how?

Should I take the returned results, somehow extract the vector and then calculate myself?

Any examples of configurations and code for working near duplicate detection are interesting.