Wiki Builder: Skill to Build LLM Knowledge Bases | Hacker News Reader

Wiki Builder: Skill to Build LLM Knowledge Bases

academy.dair.ai

134 pointsomarsar19 comments

https://github.com/microsoft/skills/tree/main/.github/plugin... is similar, works for Claude Code, too. I ported it to Pi: https://pi.dev/packages/@amb007/deep-wiki?name=deep-wiki Elsewhere, I added deep-wiki:lookup that's like deep-wiki:ask but giving precedence to the wiki instead of the code.

I’ve done this for my health records but my Claude web app (which I want to be able to access them) can’t read GitHub as effectively as it can a huge markdown file that just collects all materials into a single 15,000 line text file I reupload every week or so.

So even though I’ve done the whole wiki / knowledge base thing. The most efficient way of handing it over as a universal file to GPT and Claude for analysis.

Claude web app cannot automatically read from my github private repo. I need to attach it every time manually.

Claude Code works though. But now I can’t share this with non-technical users

Yeah I ended up just adding to claude.md for it to make a single file summary of the entire vault at any new update. It works fine.

But it would be nice to directly hook it up to a GitHub repo that I can commit and push to. I’m guessing MS won’t let them build that to safe guard their own models and OAI investiment.

At this point Claude so is far ahead in “empathetic understanding” and relational understanding, it’s painful to use GPT for anything but extremely factual content. IMO both Gemini and GPT are now overwhelmingly a failure as general purpose chatbots, while Claude is increasingly resembles what I imagined an AGI system to be.

Is this a name collision or is the related to the DAIR institute

https://en.wikipedia.org/wiki/Distributed_Artificial_Intelli...

Looking at the name that keeps showing up on the linked site, I suspect a collision is most likely (to be charitable).

I was similarly confused, and did some digging to determine that they are not related:

https://twitter.com/DAIRInstitute/status/1830719509931663411...

Maybe it is me, but I am much more interested in a way to maintaining a wiki.

Ideally something that doesn't require rescanning all the sources and something that can verify that the wiki is correct.

I am building a bit of that infrastructure for All About Berlin. There is a linter that regularly updates constants from information found on other websites. I want to keep expanding it to cover more facts, and include more elaborate source checking.

If the contributor instructions for your wiki requires:

1. forking the repo

2. committing the changes

3. submitting a pull request

... then you don't have a wiki.

I've been thinking about something in this space, actually... it feels like this is much more a UX/social problem -- in that a wiki can very much be modeled as a repo with a very permissive auto-merge bot (e.g. if PR only touches unprotected pages and user is registered, allow merge)

> it feels like this is much more a UX/social problem

It's not merely "like" that. That's what it is.

"Wiki" comes from the Hawaiian work for "quick". You spot an error, you click the button to change it, and the change is made. That's wiki.

"Open a pull request and get it approved" is not wiki. It's what the default collaboration model was before wikis and exactly why the wiki was invented (to replace it).

If the PR automatically gets approved (given checks) then pushing the merge button is just going from draft to published.

Being able to work on a draft without publishing, and incorporate changes (i.e. rebasing) should make your life easier not harder.

I find this really annoying too. A wiki is not a knowledge base. For some reason people into LLM's seem to have decided to call things wikis, I am guessing because they want the credibility wikis have.

I agree and I'm guilty of creating what is effectively a heavily hyperlinked knowledgebase and calling it a wiki. Unfortunately, only a tiny majority will ever create or edit a page despite the incredibly low barrier of a web browser without minimal authentication.

From Ward Cunningham himself:

"A wiki invites all users—not just experts—to edit any page or to create new pages within the wiki website, using only a standard 'plain-vanilla' Web browser without any extra add-ons."

"A wiki is not a carefully crafted site created by experts and professional writers and designed for casual visitors. Instead, it seeks to involve the typical visitor/user in an ongoing process of creation and collaboration that constantly changes the website landscape."

---

Wikipedia is effectively a crafted site that is maintained by experts (or at the very least very knowledgeable amateurs who 'own' certain domains) designed for casual visitors. The idea of a Wiki is great but in practice I'm less confident it exists as envisioned.

> A Claude Code Plugin

hmm interesting startup idea emerging here.

This is a really interesting direction. Thanks for sharing!

a pull request required to update a wiki? bruh.

Reply on news.ycombinator.com