What’s the correct way to chunk text for embeddings in a RAG pipeline?

python neural-network artificial-intelligence large-language-model retrieval-augmented-generation