Using gpt-4 API to Semantically Chunk Documents

[SomebodySysop]

@IntelliJJ

I am having some trouble with my current implementation that tries to do this. Mine works 90% of the time, but sometimes fails to put it back together in a valid JSON object.

I use line numbers in the original text document to always reliably retrieve all content for the final JSON output. Here is my high level diagram: