WikantikKnowledgeGraph

---\r\ncanonical_id: 01KQTCBW5GBFJVWYB8V1CP49P5\r\nverified_at: '2026-05-04T21:10:44.598011331Z'\r\nverified_by: gemini-cli-mcp-client\r\nsummary: A comprehensive guide to the Wikantik Knowledge Graph, frontmatter projection, and the proposal workflow.\r\ntype: article\r\ndate: '2026-05-04'\r\nstatus: official\r\ncluster: wikantik-platform\r\ntitle: The Wikantik Knowledge Graph\r\nrelations:\r\n- target: 01KQTCAKV3BVHYPW20PHSFGXJR\r\n relationship: part-of\r\n---\r\n# The Wikantik Knowledge Graph\r\n\r\nThe Knowledge Graph (KG) is the semantic backbone of Wikantik. It transforms flat Markdown pages into a rich, queryable network of entities and relationships.\r\n\r\n## Core Concepts\r\n\r\n### 1. Nodes and Edges\r\nThe graph is a **Property Graph** stored in PostgreSQL. Every node tracks its `source_page` and `node_type`. Edges represent typed relationships like `part-of` or `implements`.\r\n\r\n### 2. The Structural Spine (Source of Truth)\r\nWikantik uses a **Frontmatter-First** approach. The `relations:` block in a page's YAML defines its outbound edges. See **[Frontmatter Conventions](FrontmatterConventions)** for the schema and **[Markdown Links](MarkdownLinks)** for linking by ID.\r\n\r\n### 3. Provenance Model\r\nTrust is tracked via provenance tags: `human-authored`, `ai-inferred` (proposals), or `ai-reviewed` (approved and injected back to frontmatter).\r\n\r\n## The Proposal Workflow\r\n\r\nTo overcome the burden of manual metadata maintenance, Wikantik employs a **Human-in-the-Loop AI Enrichment** cycle:\r\n\r\n1. **Extraction:** An AI agent scans content. See **[Knowledge Graph Construction Pipeline](KnowledgeGraphConstructionPipeline)** for the technical stages.\r\n2. **Proposal:** Suggestions are submitted to the `proposals` table.\r\n3. **Review:** Administrators approve or modify suggestions.\r\n4. **Injection:** Approved proposals are written back to the page's YAML frontmatter.\r\n\r\n## Traversal and Tools\r\n\r\nThe KG is exposed to agents through specialized MCP tools like `traverse`, `query_nodes`, and `find_similar`. These are documented in the **[MCP Integration](McpIntegration)** guide.\r\n\r\n## Architectural Benefits\r\n\r\n- **Context Compression:** Millisecond queries for deep relationships.\r\n- **Discovery:** Finding semantically related topics without keyword matches.\r\n- **Validation:** Automated checks for prerequisites and verification status.\r\n","expectedContentHash: