Need advice on chunking strategy for RAG based OpenAI chatbot

I have built a RAG-based chatbot. I want to perform section-wise chunking on my document, which consists of data in the following order: header, description, image URL, and graph markdown. Let’s refer to this as a section. In this order, I have multiple sections in my document. I want to chunk my document so that each chunk consists of one complete section, from the header to the graph markdown. I am using a semantic splitter node parser from the Llama Index to perform the chunking.

1 Like