Prompt Palette made - should I pull request?

[samwega]

I'm entirely new to all this but keeping it short, as you can see I forked and added major new functionality in the form of a prompt palette, including both custom and default prompts, and reworked Settings page. It can now be used even for translating text, and pretty much anything people want.

I believe this diverges significantly from the intended use of Proofreader, so I haven't pull requested, but I can do that if it's desirable. Here are some screenshots of the functionality:

Settings:

[chrisgrieser]

That is pretty cool, however as you correctly noted, this is diverging pretty far from Proofreader's original goal. The main idea behind Proofreader is to use a diff algorithm to show which parts of a text exactly have been changed.

If you translate text, such an approach is not necessary, since 100% of the text is being rewritten instead of just a smaller part of it. Thus, I think this would make more sense as a separate plugin rather than an extra feature for Proofreader. (Though, I assume, creating multiple custom prompts this way is likely already available at one of the many AI-rewriting plugins.)

[samwega]

I've pretty much cloned the whole functionality of type.ai's inline diff, including their prompts full list (tho the actual prompts are secret).

Translation is just one of many possible prompts, which can be toggled off. Type.ai for instance does not offer translation built in prompts, but they offer the custom prompts which you can use in this or any way you like. You could probably even get code suggestions/corrections if that's your use case. Every prompt other than translation makes normal use of the diff. The only plugin I have seen which allows custom prompts is Obsidian Copilot (there may be others). It looks like this :

However, Copilot simply replaces the selected text, without allowing a side by side comparison (let alone diffing), so even in the case of a translation I would want to see them side by side before accepting/rejecting. I would often be translating between multiple languages I can read, thus be able to see if I like the translation before accepting is helpful.

You seem set on maintaining Proofreader very narrowly focused on the original goal (which I respect). My idea is to make it a feature complete writing assistant, which is what I was looking for in the first place—it should maintain every feature of Proofreader, and allow one to use it exactly like Proofreader if they want—or choose to do much more with it.

My philosophy is to treat the user like an adult, so maybe they want to use different models for their own reasons, they should be able to choose freely. Maybe they want to use different highlighting which breaks markdown compatibility, or maybe they want to maintain it as Proofreader does, I want them to be able to make that choice. And I want for the custom prompts to allow people to use the plugin in ways I can't even imagine right now. I think people will find new creative ways to use it that will teach me something back.

[chrisgrieser]

I have created 50+ plugins, and that experience over the years has just shown me that keeping things narrowly focussed is important to avoid things like feature creep, which at some point make your projects unmaintainable. It's also known as the UNIX philosophy: Do one thing and do it well.

(There is also the pragmatic reason that I have a full-time job, and I am creating my open source projects for free in my leisure time. My time is limited, and adding features I do not need myself effectively means that I work for other people for free on top of my job.)

Nonetheless, even if it is not about translation, almost all your prompts still result in large rewrites or additions, so diffing the changes simply does not make much sense for that kind of change. For your use case of comparing original and change, you should rather use a UI dedicated to side-by-side comparisons. However, that would diverge from Proofreader's intended goal too much, and would thus make more sense as an entirely new plugin.

My philosophy is to treat the user like an adult, so maybe they want to use different models for their own reasons, they should be able to choose freely. Maybe they want to use different highlighting which breaks markdown compatibility, or maybe they want to maintain it as Proofreader does, I want them to be able to make that choice.

Creating a focussed project does not mean treating users like children, it's a specific design philosophy which I find better than having feature creep. Nonetheless, if you are able to code, then there should be nothing stopping you from creating your own plugin where you can follow whatever design philosophy you prefer.

[samwega]

Oh, I am not able to code at all! I did a bit of very basic coding many years ago, forgot most of it. I did all this with Windsurf (GPT 4.1) and I am literally shocked it didn't turn into a steaming pile of dung. As far as I can tell, the code is pretty clean. What I can do is tinker, follow Gemini 2.5 instructions and keep learning how to use these new tools.

I understand this would be a monster to maintain, if I were to make dozens of plugins. I just really want a good writing assistant and I don't think there is one that comes even close yet, as far as obsidian plugins go. And I will have to disagree with you on the usefulness of this, after having paid to use the feature and seeing what it can do, I feel handicapped trying to create content / writing without it, it's an absolute game changer. I hardly ever used the sidebar chat, which feels like a downgrade at this point, for that type of workflow. I can always get Copilot or Scribe or some other sidebar AI to do that—I agree with the do one thing and do it well, it's just what I need to do is not exactly covered by Proofreader. For instance, Copilot does everything, but I don't like how it does most things, so it doesn't do them well as far as what I need from it. But I can use the one feature it does well, to chat with a note, and use your inline diffing for my writing purposes.

I time-stamped for you how this looks in action in Type.ai: https://youtu.be/jPb19UbQtpQ?t=172

Check out what I was able to do here as another example:

Prompt used: You are an expert writing assistant. Refine the structure of the following text, in ways compatible with Obsidian Markdown (#headers, - bullet and 1. numbered lists, **bold**, *italics*, etc.) Make sure it makes sense, and don't overdo it. Only output the revised text. This is the text:

The multi-line diffing looks funny but I guess I will have to fix that next. I suppose Proofreader's logic is not designed to work across multiple paragraphs? (or maybe I broke something)

Anyway, I posted here because I wanted to make sure whether I should just fork hard and go in this new direction freely, or if I should maintain strong compatibility with Proofreader in order to pull request and help with its development. I am taking your response as permission to do my own thing with it.

[chrisgrieser]

As per the MIT license of the project, you never needed permission from me anyway. Just credit me, if you end up using parts of my code, then it's all good with me.

And once again, I am not saying that what you suggest is not useful. Just that I prefer to keep a narrow focus on my plugin, and that I do not have the time to work for free adding features for others that I don't need myself.