Is there a retrieval only advance

The community members are discussing how to parse Markdown files while preserving their structure, and use post-processors to include adjacent nodes before parsing the prompt. They are looking for a tutorial or example on how to create a graph structure with headers as parents and paragraphs as child nodes. The discussion also touches on the structure of the llama-index library, with one community member suggesting that retrieval should be a separate section from the query module, and that the query and chat engines should be built on top of the retriever instead of the index. The community members provide some resources, such as a custom Markdown parser and a low-level API example, but there is no explicitly marked answer to the original question.