JsonConverter for Converting JSON Files into Structured Markdown Files

[ZeyuTeng96]

This PR implements a JsonConverter class that converts JSON files into Markdown files while preserving the dictionary structure of the data to the greatest extent possible during the conversion process.

• For JSON files with dict like:

The conversion result will be:

• For JSON files containing a list of dicts, such as the SFT fine-tuning data for LLMs:
  

The conversion result will display each element of the list in sequence. Additionally, users can customize the value of the list heading (default is Elem):

This conversion approach preserves the dictionary structure to the greatest extent possible, allowing converted .md files to be more intuitively readable and visually clear for users. Furthermore, when performing retrieval-augmented generation (RAG), LLMs are highly familiar with dict structure and can process them effectively.

[ZeyuTeng96]

@microsoft-github-policy-service agree

[afourney]

Thanks for the PR! We appreciate it!

Both #219 and this PR propose alternative ways of converting JSON to Markdown.
An alternative approach would also be to simply quote the json (perhaps pretty-printed) in a Markdown code block.

Ideally, we would have a clean way of registering/unregistering different converters, so that people can customize the library to their desire (e.g., by picking from the 3 above-mentioned approaches). I think this is coming in the new year, and will discuss with my team when everyone is back.

Until then, I'm not sure I am ready to pick a default behavior yet (beyond treating it as plain text -- which is what was intended)