I have been using SB for a few days, setting up my own PKMS. It's been fun. I added the AI plug and have found success with the free tier Gemini API. However, there's quite the issue with LLMs (Gemini anyday) getting confused between v1 and V2 Syntax. I can see why given how the documentation is laid out and available online. Has anyone experienced this or found a solution other than writing code for queries etc yourself? Is it something the team have thought about in terms of documentation?
I haven't thought about it explicitly, but it's an interesting issue. Similar to SEO before it, it's not super clear how AI companies index the internet and how often they do it. I could take v1.silverbullet.md offline (and maybe at this point I should, it's been a long time and I doubt many people actively use either v1 or at least the site anymore), but I have no idea how long it would take for this content then to disappear from LLMs (if ever). I may do this actually. Then there's still a bunch of v1 content on the community that is exposed. I've already disabled including results from this category from the community's search engine, but perhaps I should just archive/delete that whole v1 legacy category 
I guess I don't know the need for a v1 resource. It's the combination of a niche language and not much online about it maybe? I did some comparison: other LLMs are a little better but only the top end Claude ones seem to be able to get it right first time.
As someone who does code, I have found space lua a bit of a learning curve. I will stick at trying to learn it as my first option 
Maybe just put an admonition on all v1 documentation stating that the documentation is for v1 version.