Blumentals Software Community

It probably does not open because you have "Synchronize with cursor" on, which selects what you click and scrolls to it in code. Uncheck that.
Like this:

This is a maintenance release.

* Various bugfixes and tweaks: losing editor focus, SFTP speed, AI history issues, +more

Excellent detail! I was able to reproduce the issue with no files open, so I will be able to fix it.

Close the editor, go to C:\Program Files (x86)\WeBuilder 2025\data\mldefs and delete HTML 4.01.ini.
I'll write this bug down but I haven't been able to reproduce it yet, maybe I'll figure it out later.

Maybe at some point, but it will require to adapt the app, as server-side is also currently doing some lifting, that would need to be moved to the application. We will probably do this once the open LLMs get as good or almost as good at coding as the big 3 of the commercial LLMs.

To change it click menu Options -> Preferences -> Language Options -> HTML and change HTML Language Specification to HTML 5.

Pushing it to the limit, I respect that. If a model wants to overwrite the entire file using a tool, unfortunately tool responses cannot be streamed, so the model just works silently for a long time until it is ready to return the data. What I have done now, I increased the timeouts so it does not ...

Actually I think I know what might be the issue. Are you trying to rewrite the entire file? Writing has more limits, you can do it but tell it to rewrite it in chunks, around 500 lines each, not all at once.

We haven't added our own extra restrictions, the models have their own limits. But I want to understand this clearly, if you have only hint about conversation getting long, despite the hint it should still return the answer and go on, or error , where the model returns error and stops. I'm looking ...

The warning about conversation getting long does not mean you have to summarize immediately, it means that the context is huge and thus each response is costly, but this is expected for large files. When you summarize the conversation, it loses the details, it knows about the file but not its ...

It's probably some other CSS from the design interfering. If you can point me to the website, I could take a look and see if some additional CSS might help with this.

My opinion is that creating a website is not the best use of AI. The reason is that when you create a website, you usually care a lot about the design and all the tiny details on the frontend, so you probably want to have hands-on approach at least for that part. My opinion is that AI works much ...

Indeed, inspector is another great way to do it.

The way to do it with minimal typing is this: write space after <p, wait for auto-complete, doubleclick on "class", wait for autocomplete, doubleclick on the class name. I'm including the screenshots. Step 1: Ekrānuzņēmums 2025-12-03 143724.png double click class Step 2: Ekrānuzņēmums 2025-12-03 1438 ...

You can download and test the newest version. It hasn't been announced yet but it is available to download already.