This would add the possibility of using local embeddings that are much faster and precise than Open AI, like nomic-embed-text. They are also free.