knowledge_web

Web documentation and articles included in my knowledge base.

Every directory here is one mirrored site, imported by kb web from the knowledge_mcp builder and indexed as part of the public collection. The knowledge base answers with the live URL first; this repo is what stands behind that link when the site moves, changes or disappears.

Layout

docs.yoctoproject.org/
  _site.yaml                        # the site: licence, attribution, crawl config
  dev-manual/common-tasks.md        # the archive — the page's text as Markdown
  dev-manual/common-tasks.yaml      # the sidecar — source URL, date, content hash
  dev-manual/common-tasks.print.pdf # the proof — the page as it was rendered

File	What it is
_site.yaml	One per site, read and approved by a human. Nothing under it is indexed until it says reviewed: true.
*.md	The page's text, boilerplate stripped. This is what gets indexed. It is an archive of the text, not a pixel-faithful snapshot.
*.yaml	url, retrieved, content_sha, title, licence. The url is what a citation points at.
*.print.pdf	The same page printed by headless Chromium on the day it was read. Stored in Git LFS, never indexed.

Reading a proof

The PDFs are LFS objects, so a plain clone gives you pointers:

git lfs pull --include "docs.yoctoproject.org/dev-manual/common-tasks.print.pdf"

Pull them only when you actually want one — the whole set is large.

Licensing

Nothing here is my own writing. Each site's licence and attribution are recorded in its _site.yaml and repeated in the header of every archived page. A site whose terms do not allow it is not mirrored; a site that allows mirroring but not machine processing is kept here with license: no-ai and left out of the index.

If you own a page mirrored here and want it removed, open an issue.