Still, I would really like to know how you approached this from an architecture perspective. I'm also curious to how much coding agents structured the code like so by themselves or if you had to steer them heavily (I've just tried a bit gemini flash from antigravity cli so I'm a bit behind on this).
Also how did you approach the rendering correspondence with actual tikz code? Do you have some tests that like render the tikz using latex and your js pipeline and compare the result for differences?
However, the generated TikZ code is not good in my opinion. Everything uses absolute coordinates, which in TikZ is seldom needed.
Just to start, if I place a single node I get absolute coordinates for it. Why? If you just write `\node {Hello};`, TikZ will put that at the center of the bounding box. No need to tell it's at `(0.5,2.91)` like it's happening in my test. Then features such as "align bottom" for a selection of multiple nodes should are manipulating the absolute coordinates instead of using TikZ's alignment features (anchors etc.).
I understand generating such code is more difficult. Maybe it can be something to point at for the next version, who knows...
\begin{tikzpicture}
\draw (0,2) rectangle (1,1);
\draw (1.5,2) rectangle (2.5,1);
\end{tikzpicture}
Elsewhere in this discussion, we talked about positioning like "right of", and some good suggestions were made (https://news.ycombinator.com/item?id=48647683).
I know people like it, but I hate writing TikZ manually, to the point that I've mostly moved most of my technical-ish drawings to draw.io/diagrams.net, and then just export to a PNG. I feel like it's inelegant, but it works well enough and it's easy to make something that looks ok. Generally I'm all for text-defined stuff.
I have moved some of my stuff to Mermaid when I know my stuff is going to live in Markdown but I've not tried to get that working in TeX.
That said, I would like to use TikZ just because it's kind of the idiomatic way of doing diagrams in LaTeX, so a WYSIWYG might be useful.
One suggestion, I would like the arrows to be able to "attach" to the boxes, as in the arrow endpoints can move when you move the boxes. That's how draw.io does it.
When using draw.io I’d suggest exporting to PDF instead of PNG so you keep it as vector graphics.
I had trouble getting that working (admittedly years ago) and as long as you have a high enough resolution people can't really tell a difference between it and SVG, though obviously it will make the filesize bigger.
Just tried the text nodes and indeed the arrows work. I guess I would also suggest doing the same for regular shapes.
I've done diagrams with it a few times and just used Computer Modern to make the diagrams look the same as the text. Good enough.
https://ctan.org/pkg/circuitikz?lang=en
https://github.com/circuitikz/circuitikz
Some years ago I wired it up with `asciidoctor-diagram` so we could have simple circuits in our Asciidoc maintenance manuals. The techs loved the hell out of it, and we could collaborate on the things in a git versioned ecosystem vs whatever fresh hell the PDM/ERP had for us.
A very nice complement to the already awesome WireViz (https://github.com/wireviz/WireViz)
How hard would it be to support cetz? I'm not touching LaTeX if I can avoid it, but I'm using Typst all the time.
There is a full wysiwyg (vibe-coded) presentation software based on typst available which partially implements exactly that:
But, I've taken my papers to Typst. Could you have the agent do the same thing for Cetz, the TikZ equivalent for Typst?
I'm looking to write a paper on a recent project, but most of the places I've seen to submit has asked for TeX. I greatly prefer Typst because of the ridiculously fast compilation times but I haven't used it for anything outside of school assignments do to that restriction.
The world thanks you.
I was quite proud of the hours of work I had put in to configure it just so, with the 3d look and all.
I'm not an AI evangelist, but this kind of thing is such a welcome boon. More itches can be scratched!
\foreach \i in {1,...,5} {
\node[circle, draw] (n\i) at ({90 - 72*(\i-1)}:1cm) {$\i$};
}
Then the user can do one of two things. (1) Select an item and place it on the grid, and the item gets replicated on all the grid points. (2) Pick and place different items on each of the grid points.
> TikZ’s name is intended to warn people that TikZ is not a program that you can use to draw graphics with your mouse or tablet. Rather, it is more like a “graphics language”.
While making the app I was worried that I was going against the TikZ philosophy. Maybe I should have named it "TikZ ist doch ein Zeichenprogramm" (TideZ)..
At some point the people who seethe with hate for AI, and claim it's all hallucinations and illegitimate hype, are going to have to admit they were wrong. Projects like this are the proof staring them right in the face, if they care to look.
(Not on HN but I do still see some folks who last tested LLMs before Nov ‘25, those folks might still be mostly out of touch.)
\begin{tikzpicture}
\node[draw] (A) at (0,0) {A};
\node[draw, right of=A] (B) {B};
\end{tikzpicture} 1. Right-click on an existing object, offer drop-down context menu.
2. Menu item `Position relative to...`.
3. The cursor now selects _other_ objects in the field.
a. If there is no other object, then offer to create a new label-less node with (x,y); default to the origin.
b. Once an object is selected, then offer `right of`, `left of`, `north of`, `south of`, `southeast of`, etc as a drop-down menu, and a field for radial displacement.
i. As a stretch goal, offer a `Custom position...` button to specify an (x, y) displacement, or a polar angle and radial displacement. These three options (fixed offsets, Cartesian, polar) could also be tabs in the resultant menu from (b) above.