The secret sauce behind LLMs are embedding vectors — just arrays of very small numbers. But with them, you can build absurdly useful things ranging from cross-language search features to polyglot chat bots. We'll talk about how they work and how you can use them to build magical things.