Just in case someone is playing with loading the markdown from a file instead of from a text variable in the Document splitting section (Context aware splitting):
The Markdown loader in langChain (UnstructuredMarkdownLoader) removes the markdown characters needed for splitting the text in the example (e.g.: #, ##, ###). So the splitting does not work.
Instead use the plain text loader, which will load the file as is and not remove anything.
loader = TextLoader(path)
data = loader.load()
markdown_document = data[0].page_content