The sitemap is intentionally smaller than the file tree. It includes the pages that represent the notebook and leaves out temporary files, implementation details and generated fallback material.
A compact sitemap makes review practical. If a new public page is important, it should appear in navigation, in the sitemap and in the JSON metadata; if it does not meet that bar, it can remain an internal asset.
This boundary keeps the site predictable for readers and for crawler tools. The visible information architecture and the machine-readable map stay aligned.
Internal links are not only navigation. They are also a basic quality check: a page that cannot be reached from related material is probably not ready to be considered public.
The notebook links article pages to their matching JSON files where that makes sense. This gives the human page and the machine-readable file the same topic, but keeps the formats separate.
The result is simple to test. A crawler can follow the navigation, a person can read the articles, and a script can inspect the JSON without needing a separate map.
The metadata policy is simple: title, description and canonical URL should agree with the visible purpose of the page.
Short descriptions are easier to audit than elaborate search snippets. They also reduce the chance that a stale page summary promises content that is no longer present.
The public metadata resources repeat the same basic facts in a machine-readable form, so a human page and an automated check can compare the same expected shape.
