Have you ever thought about introducing ChatGPT as a document generation engine, just like Mintify Writer?

[mkdir700]

like this: https://github.com/mintlify/writer

[ColinKennedy]

This isn't my repository so it isn't my place to say but for what it's worth, docstrings (documentation) are for contextualizing why X or Y function / class is important or point out any risks or input assumptions, etc, or just reasons to use the function.

The linked demo shows auto-created docstrings that simply restate what the code is already doing. A docstring like that is something that you could learn just from reading the source code, itself. IMO such docs are worse than none at all because at least if a docstring is missing, linters like pydocstyle can point out where documentation is lacking so programmers can go back in and improve them.

It'd be a different story if the auto-generated text gave actual insights about the code though. Since neogen uses snippet expansion engines, I could see a use-case where, say, you have a function like:

def foo(bar, fizz, buzz=tuple()):
    """Do blah, I am a description, written by a human.

    Args:
        bar (SomeClass): |Your cursor is here|
        fizz ([TODO:parameter]): [TODO:description]  
        buzz ([TODO:parameter]): [TODO:description]
    """
    for item in buzz:
        # ...

And, using an AI, give it an automated prompt like "given 1. the entire function body and 2. the manually written description, "Do blah, I am a description, written by a human." and 3. SomeClass type, suggest a description for the fizz parameter because that's where my cursor is currently positioned"

A thing like that might be useful. But anyway, this is borderline AGI territory, IMO.