💭 LLM now provides tools for working with embeddings