[bootsmann]

You serve the embedding model in a lambda and then run something like FAISS in the backend.