Recommend similar documents in your dataset, based on vector fields. You can feed this endpoint a list of 'positive' or 'negative' _ids for documents that you want to base the recommendations off.
You are even able to facet and filter your recommendations, like with search. However, note that the syntax for filtering recommendations is different (for now) to search. Please visit this page for details: https://discovery.relevance.ai/reference/filtering-recommendations