Knowledge Graph

The knowledge graph is a visual inspection surface for wiki connectivity, not the source of truth.

The graph visualizes pages and cross-references built from the same wiki index that powers search and navigation. It helps agents and Kevin spot isolated pages, dense hubs, and unexpected clusters. Source: UI implementation; Backlink Index Protocol

What It Shows

Element Meaning
Node Wiki page
Edge Wikilink, related frontmatter, typed relationship, or tag-derived connection
Node size Connection count / centrality
Edge strength Link frequency or relationship weight
Filtered nodes Private/gated categories hidden by server-side visibility rules

How It Is Built

  1. npx tsx scripts/build-index.ts regenerates page metadata and backlinks.
  2. The UI reads the page index.
  3. The graph builds nodes from visible pages.
  4. Edges come from wikilinks, frontmatter, and shared metadata.

There is no separate graph export step. If the graph looks stale, rebuild the index first.

Agent Uses

  • Find orphan pages after a capture or migration.
  • Spot over-broad hubs that may need splitting.
  • Inspect whether new pages connected to the expected cluster.
  • Verify that important meta pages point to each other.
  • Explain wiki structure to Kevin when the graph itself is the topic.

Interaction

Action Result
Hover title, category, connection count
Click camera focuses on node
Right-click open article
Controls auto-rotate, reset, labels

Maintenance

Graph quality depends on link quality. Do not "fix the graph" by adding meaningless links. Fix the underlying pages: related frontmatter, inline wikilinks, and entity backlinks. Then rebuild the derived graph surfaces through Generated Surface Contract.


Timeline