There is this research from ChromaDB that talks about the different strategies for chunking text without loosing valuable context for optimal RAG systems, but I'd recommend checking out Adam Lucek's YouTube channel where he demonstrates with practical examples these strategies. He also made available a Github repository (QuicKB) where he puts in practice all this. I think it's a really valuable resource. Don't miss the link to the video below! Cheers! 🙂