Skip to content

Knowledge Graph

The Knowledge Graph is Baan’s view into the conceptual structure of your content. It can extract entities (people, organizations, places, concepts, events, products, creative works) from published articles, link them to Wikidata IDs, and render the result as an interactive graph.

This is a rare CMS combination: entity extraction, Wikidata disambiguation, graph visualization, and AI-driven article suggestion in a single tool.

When entity extraction runs, Baan asks the configured AI provider to identify named entities and classify them into one of seven types:

TypeExamples
PersonSteve Jobs, Ada Lovelace
OrganizationVercel, OpenAI
PlaceTokyo, Silicon Valley
ConceptServer Components, Functional Programming
EventWWDC, Apollo 11
ProductNext.js, Claude Code
CreativeWorkBooks, films, papers

Extracted entities are stored in the post’s frontmatter and indexed in the article_index.entities column for fast graph queries.

The same name can mean different things in different contexts. “Apple” might be the fruit or the company; “Java” might be the language or the island. Baan can resolve this ambiguity by attaching a Wikidata ID to entities:

  • Apple (the company) → Q312
  • Apple (the fruit) → Q89
  • Java (the language) → Q251
  • Java (the island) → Q3757

Two articles that mention “Apple” in the company sense are linked in the graph; an article about apples (fruit) is not. This drastically reduces false connections.

The Wikidata ID is also emitted in the article’s JSON-LD mentions and about properties, helping AI engines and search crawlers correctly identify what the article is about.

Every extracted entity becomes a node. Articles that share an entity are linked. The graph is interactive — pan, zoom, click a node to see all articles tagged with it.

The visualization makes content gaps visible at a glance:

  • A dense Next.js × React cluster suggests strong coverage of frontend topics
  • An isolated UI Design node suggests you’ve mentioned it but never written deeply about it
  • A heavy Vercel node next to a thin AWS node reveals which platforms you’ve actually engaged with

Select two entities on the graph. The Bridge Article Generator prompts the AI to produce 5 article ideas that connect those two entities, each with:

  • A working title
  • The angle / hook
  • A complete outline

Click “Create Post” to land in the editor with the title, entities (already in frontmatter), and outline pre-filled. You write the body — the structural work is done.

This turns the Knowledge Graph from a passive visualization into an active content discovery tool. Instead of staring at a blank page, you start from “what can I write that bridges A and B?”.

  • Entity extraction runs on the configured AI provider (Anthropic / OpenAI / Gemini). Set the provider under Settings → AI.
  • Existing articles can be re-indexed in bulk via the Settings UI or POST /api/cms/v1/index/sync.
  • The full spec lives at docs/spec/knowledge-graph.md in the source repository.