A fork of Calibre called Clbre, because the AI is stripped out | Hacker News Reader

A fork of Calibre called Clbre, because the AI is stripped out

github.com

28 pointspabs324 comments

Open articleView on HN

The AI integration in question, from the Calibre changelog:

- Allow asking AI questions about any book in your calibre library.

- Right click the "View" button and choose "Discuss selected book(s) with AI"

- AI: Allow asking AI what book to read next by right clicking on a book and using the "Similar books" menu

- AI: Add a new backend for "LM Studio" which allows running various AI models locally

It seems pretty harmless really.

I understand some people feel that AI is overhyped and don't particularly like it, but this level of weird knee-jerk "anything AI is the devil incarnate" response is just as ridiculous, IMO.

Recommending books based on reading history is one of the very few uses of AI I'd consider good. With enough metadata about books it will be able to recommend similarities better than anyone, with appropriate topical filters. Storygraph is a Goodreads alternative trying to make use of this.

It would probably be better to use a deterministic graph algorithm to do this, but it's seemingly too hard for anyone to do properly.

If anything, it took time away from improving Calibre's UI and user-friendliness. I use it weekly, but by the gods, it's neither pretty nor friendly.

Calibre has had a comically obtuse and amateur UI for over a decade. I don't think anyone working on it is spending time on a grand beautiful UI redesign (which would be really hard, both in effort and in politics). So I don't think it's fair to complain that anytime someone adds a feature you don't use that is taking time away from UI work.

I wasn't complaining, I don't mind whichever feature an open-source freely-available project I use weekly wants to work on. I did suggest these features might have taken some effort, and opportunity cost dictates it wasn't spent elsewhere.

Does it make use of LM studio for all of those features? I feel the features can simply be ignored by anyone who doesn't want to use them.

From Calibre's repository README:

> Supports hundreds of AI models via Providers [...] no AI related code is even loaded until you configure an AI provider.

This fork is pretty much useless.

That's not true: there's some menu items and supporting code by default.

Still, the menu item is not interacting with AI without you explicit configuring it.

I bet if you click it without any configuration will give you an error message.

How many inactive menu items that error out when clicked on is acceptable? Are we ok with a Microsoft Word style ribbon of controls that do nothing?

If UI bugs are really the issue, then one just sends patches to the upstream project - I'm sure the maintainers will be happy to receive fixes for broken menus. A fork for this is useless, and guaranteed to be abandoned.

Not just useless but another fork that only confounds newcomers and users looking for help. We can be generally opposed to AI without making it a boogeyman.

Post this when a release is made. Currently this fork doesn't have any commit (beyond stating the intention of the fork without explaining why) and it is useless.

As best I can tell Goyal started adding AI-related code to Calibre back in August, merging the LLM tab work from https://github.com/kovidgoyal/calibre/pull/2838, and created the chat widget in November with commit 41d0da4267dc6f7f7e48fb9bb7e8609a2e251cb7.

I looked at forking the project myself: the challenges are that it's a very quirky application, its design and implementation doesn't share conventions with any other application, and the build system is complex and unique to Calibre.

It's a shame there's no good open source ebook library application with a more conventional design. Shoving AI into everything, even when it defaults to "off" (for now), is getting old.

I personally think Goyal should focus on making Calibre’s UI less unpleasant before adding yet another bolted-on AI chat textbox, but this doesn’t feel very productive or useful.

"Unpleasant" is a fairly good word for describing Calibre's UI (as much as I love Calibre)

It's not terrible, but it's not great. You get used to it very quickly, but it's still clunky.

Oh well. I suspect that sort of update would be a lot of refactoring. Supremely happy with Calibre altogether :)

The only difference is the README.

I'd prefer a fork that uses AI to convert ebooks into custom audio books.

It’s 1 readme commit. It should say, “an aspiration to modify Calibre”

I haven't use Calibre in several years. What does it use AI for?

https://calibre-ebook.com/whats-new

> Allow asking AI questions about any book in your calibre library. Right click the "View" button and choose "Discuss selected book(s) with AI"

> AI: Allow asking AI what book to read next by right clicking on a book and using the "Similar books" menu

> AI: Add a new backend for "LM Studio" which allows running various AI models locally

Looks like it's pretty standard: ask AI questions about your current book and across all books in your library.

I had to look into the changelog to see what they meant but it’s not too intrusive IMO [1]. That being said, I find it way easier using calibre-web-automated [2]. Everything is through a web server, has usual control, and syncs with Kobo. All it takes in one docker compose and you’re up in less than 10 minutes.

[1] https://calibre-ebook.com/whats-new [2] https://github.com/crocodilestick/Calibre-Web-Automated

A good usecase for LLMs with an option to use local models?

Better remove it.

Reply on news.ycombinator.com