I finally finished up my post on Bibilobunny, my book note extraction tool for Kindle and Play Books:

https://battlepenguin.com/tech/...

I hope to add support for getting notes off Nook and Kobo next. You can follow the instructions to create your own book quote bot, and you can follow mine here: https://tweeflood.com/@bookquotebot

Cool! I gonna PEP8 you! (do you know that from the comedian? :P)

I've skimmed your source. Nice stuff.

I only have only found this with the quick few:

- I'm a huge fan of sqlite (even read the source code) / SQL too but you should really consider to use dataset. It has so much benefits. You can do db['non-existing-table'].insert(name="Book",chapters=42,pages=420) and it will create the table for you with those specific data types. It's a wrapper around sqlite.

- pathlib.Path is nicer than all those os functions like join etc. With pathlib it's just pathlib.Path(".").joinpath("pony").rglob("*") and stuff.

- for file open and stuff I would also use pathlib

But nice usage of click and no useless asyncio shit. Nicer than most stuff I make.

If you're interested I can run the AI review tool over your source like I did with Lensflare's code. It gives a grade per file. I average at a six or seven. Rarely an eight. I focus bit more on delivery than quality.

@retoor Thanks for the code review.

I use Pathlib some places in some code, but not as a full on replacement for os stuff. Might look into that more.

Yea, async is confusing as fuck. It doesn't work the way Futures/Promises work in most other sane languages.

Although this tool is simple, other tools I make need migrations to the database for newer releases. Plus, both Gentoo and Void Linux package yoyo-migrations, but they don't seem to package dataset. Although I won't package bibilobunny, I have two other projects I'm planning on packaging for other distros.

No thanks on the AI. I've found most coding LLMs are good for wrong answers only or spewing out unreadable buggy bullshit. Next token prediction models aren't the best thing for novel, complex problems.

@djsumdog Soon you can review my snek.molodetz.nl source if it's finished. It's now in full development. The source is fine, but you'll see that I do less effort than you to make the code so shiny. I decided that code should be a 6 or 7, maybe 8. But higher than that is a waste of time imho. It's just a choice which is fine to do in private projects. Got a bit traumatized at an employer that delivered nearly nothing because it kept on refactoring the F out of everything. The rule was to refactor everything what you touch to the current standards. It kept changing. Quite complex hosting provision software. Thousands of tests. Every damn TIME :P

AI not for complex novel problems, but for just application building with models and forms it outshines. It knows all my names from python model, applied it to javascript and saw the field names of other html components and made this single perfect tab complete. This is codeium. The innerHTML vs innerText is my buggy btw.