I suggest to read the documentation. In any case llm predictor should be the one through you call the llm, the embeddings are a representation of your text in vector form…for embeddings the best practice is to use ada-embeddings-002, for llm text-davinci-003
