improve the parsing when indexing

[mongodben]

nothing specific in mind here, but i imagine the parsing could be improved to improve answer quality.

couple of ideas:

1. right now it's parsing text to some 'pseudo-markdown'. it works pretty well. however i think if the html were parsed into real markdown, the results would be better. especially the code example.
2. explore using some langchain utils for parsing. they have some more advanced stuff https://langchain.readthedocs.io/en/latest/modules/indexes/examples/textsplitter.html?highlight=markdown

[mongodben]

blog posts i mentioned in meeting:

• https://dagster.io/blog/chatgpt-langchain
• https://supabase.com/blog/chatgpt-supabase-docs