This week in FreeCAD development:
Sketcher
- theo-vt prevented self-tangency from happening at the command level (PR#31042).
- YashSuthar983 patched sketch visualization to have two configurable parameters: transparency of the overall axis and occluded transparency for the occluded axis, so that an occluded axis can be seen (PR#26594).
- theosib fixed auto-constraining to external geometry in certain scenarios (PR#30970).
Assembly
- LubuSeb changed Assembly activation to use the same scoped recompute path used by the manual Solve Assembly command, instead of calling the solver directly from edit-mode activation (PR#30744). This fixes a bug where activating an assembly would block automatic recomputation after changing the VarSet parameters.
- PaddleStroke fixed incorrectly placed newly created joints (PR#30669).
- Lgt2x switched the Assembly workbench from OpenCV to PyAV to reduce the size of the builds without any functionality loss (PR#30333).
TechDraw
Thanks to pavltom, now you can attach any expression supported by the FreeCAD expression editor to any of the TechDraw template’s fields (PR#27117).

CAM
- tarman3 fixed multi-target drilling in Helix (PR#30611) and patched recomputes in SelectLoop (PR#30908).
- jffmichi removed unneeded recomputes that would happen every time a simulation was started (PR#30927).
- dbtayl fixed path generation failures in Adaptive (PR#30623).
- davidgilkaufman fixed a bug where Adaptive clearing would go around the workpiece perimeter when “Finishing profile” is selected (PR#30831).
FEM
- marioalexis84 set the default pressure value to 1 MPa (PR#31036).
- wmayer implemented methods to get element colors (PR#30726, cherry-picked by maxwxyz).
Other changes
- Roy_043 added support for multiple objects in hatches in Draft (PR#30578).
- tritao added view rotation by dragging the NaviCube (PR#30727).
- vsergeev added support for vertical configuration tables in Spreadsheet (PR#29656).

Additional improvements and fixes were contributed by Amine0x0, awgrover, chennes, eeshsaxena, greg19, Lgt2x, Roy-043, tarman3, theosib, tritao, and wmayer.
If you are interested in testing the latest weekly build, you can grab it here.
Activity stats
PR stats: since the previous report, 41 pull requests have been merged, and 49 new pull requests have been opened.
Issue stats: overall, there are 3495 open issues in the tracker, up by 17 from last week.
Elsewhere in the community
Ephi Blanshey recently shared his progress working on the History workbench that helps version-control changes in FreeCAD projects natively and visualize differences between model revisions.



6 responses to “WIP Wednesday, 1 July 2026”
The history workbench fills a need I’ve had for a long time. I am a bit saddened to see that it’s vibecoded, however. Of course, not a big deal for a proof of concept, but with that in mind, is there any potential plans to develop a built-in version control functionality? Perhaps inspired by said workbench, and taking into avvount lessons learned in terms of UI design and ergonomics. But clean-room designed to avoid the ethics issues that comes with the AI coded one.
I’m having trouble following your anti-AI slant here. A solo developer decides to use tools available to them to make something nice for everyone. Gives it away for free, and asks for nothing in return, and you’re “saddened” and hoping someone else steals their idea and builds it without AI? When mega-corps like Microsoft or Google leverage their enormous capital to do this, we do not call it ethical. Depending on the model used here, the company that trained the model may have acted unethically in the past and/or present, but that nuance is completely absent in your comment. An end user trying to improve others’ lives doesn’t seem unethical to me at all. How about we fix copyright law first, then apply it equally to everyone, and even punish the ones that broke even the old law, and perhaps we then start building more ethical AI? Sounds great to me, but since we’re all pretending OpenAI and the rest didn’t steal their training material, let’s not crap on the people who use AI to learn, build, share, or even try to improve their own miserable lives.
I don’t see any mention of this being vibe coded anywhere. On what basis do you attack this workbench for being vibe coded? Got any links or are you just going off of vibes?
Here you go: https://github.com/eblanshey/HistoryWorkbench/blob/b78e2d97db8567f08fd07df88277227036427c7d/AGENTS.md
First off, I hadn’t found the updated FreeCAD policy on LLM use until just after making my post, so that’s on me. However, I follow some PRs and issues over on the Github page and according to what was said there, there was previous a blanket ban on all LLM use. So that gave me the impression that a project like this couldn’t ever be merged as-is, which is why I brought up reimplementation. I think VCS is so central to a fully featured CAD suite that it shouldn’t have to be done by a plugin.
With the updated policy and the “you’re liable for the code you generate” clause, that might not be necessary. Though I would be hesitant to merge anything of this scale (this repo is fairly large for a plugin) into the main project, even if it was created by traditional means.
—
You have a lot of points, some addressing things I never said. At least I don’t think I did? Since this is a blog post comment chain and not a proper forum discussion, I don’t see the point of making it into a huge back and forth debate. But if you really want to engage with the discussion and where I am coming from, keep reading.
I never said anything about “stealing ideas”, in fact, I praised the project and suggested that it should be incorporated in the main program as a clean-room reimplementation exactly to avoid those kinds of concerns. The same way that open source development has always built upon and improved upon unethical closed source ideas. And the assumption is that this should be done in collaboration with the original uploader, and crediting their work. Like academic endeavors have always been done. That’s why I referred to the third party project as a proof-of-concept…
I don’t understand where you got the idea that I am excusing the unethical behavior of the megacorps that trained the models? Quite the opposite? It’s the core of almost all anti-genAI discussions, and one of the big two reasons why people are boycotting things using it. This has been discussed to death in the tech community, so I didn’t pepper my post with caveats to avoid stepping on billionaire toes.
Maybe we frequent different communities, but this sentiment isn’t in any way radical. And yes, the rampant acceptance of vibing in the open source world does sadden me. A whole lot. It mirrors the exact same trend from the tech industry that has lost both me and my close friends their livelihood. And puts fuel on the Brandolini’s law-shaped fire that threatens my new career as a teacher.
You’re free to make your case that there will some day be an ethical AI (I disagree), but even you agree that we aren’t there yet.
One of the things we as open source contributors have, that industry projects do not, is time. Proof-of-concepts and licensing workarounds create tech debt. But unlike for-profit shops, there is both plenty of time and volunteers to turn them into unproblematic, clean, features that can be mainlined. Obviously the ball wouldn’t roll anywhere without anyone willing to make a PoC. Now they have, so we shouldn’t be stuck arguing about how it can never be improved upon, and digging our heels into the ground for ideological reasons when the suggestion comes up.
Better is better. Better is never the prototype.
But nobody is proposing this workbench for inclusion in FreeCAD yet
Absolutely, I understand that! Except me, I guess: I was trying to pitch the idea, since I think this has the potential to grow into something great that ought to be included out-of-the-box 🙂