Knowledge Graph Builder
Organize information as interconnected knowledge graphs.
Usage
- Define the scope: personal notes, team wiki, project documentation, research corpus
- Choose your tool based on scale: Obsidian (personal, file-based), Notion (team, structured), Neo4j (large-scale, query-able)
- Establish node types (concepts, people, projects, resources) and relationship types (relates-to, depends-on, created-by)
- Build incrementally: add nodes as you encounter information, link to existing nodes
- Review and prune regularly to keep the graph useful
Examples
- Personal Zettelkasten in Obsidian: Each note is one atomic idea (not a topic dump). Notes link to related notes using [[wiki-links]]. Structure: fleeting notes (inbox) → literature notes (from reading) → permanent notes (your own thinking). Over time, clusters emerge that reveal unexpected connections. Tags for retrieval, links for discovery
- Project knowledge base: Nodes: requirements, decisions, people, systems, meetings. Relationships: "Decision X was made in Meeting Y by Person Z regarding Requirement W." When someone asks "Why did we choose PostgreSQL?", you can trace: Decision → Meeting minutes → Discussion context → Alternatives considered. Institutional memory preserved
- Research literature graph: Each paper is a node. Relationships: cites, contradicts, extends, replicates. Author nodes connect to paper nodes. Methodology nodes connect papers using similar methods. Querying: "Show me all papers that contradict Smith 2024" or "What methods have been used to study X?" Reveals gaps and opportunities no reading list can
Guidelines
- Start small: 10 well-connected notes are more valuable than 100 isolated ones. Quality of connections > quantity of nodes
- Write in your own words, not quotes — the act of rephrasing forces understanding and creates more meaningful connections
- Link generously: every new note should link to at least 2-3 existing notes. If it doesn't connect to anything, it's either the start of a new cluster or not worth keeping
- Use a consistent naming convention: "concept-name" for ideas, "person-firstname-lastname" for people, "YYYY-MM-DD-topic" for meeting notes
- Review your graph monthly: are there orphan nodes? Dead-end clusters? Over-connected hubs that should be split? Prune and restructure
- The graph is a thinking tool, not just storage — browse connections to generate new ideas and find non-obvious relationships