@wiki-nexus/core

Core indexing engine for WikiNexus — parses markdown wikis into knowledge graphs stored in SQLite.

What it does

• Parses markdown files with YAML frontmatter, [[wikilink]] extraction, and heading hierarchy
• Builds a relational knowledge graph in SQLite (pages, edges, tags, sources)
• Clusters pages into topic communities using Louvain community detection
• Traces knowledge flows from source documents through derived pages
• Analyzes impact (blast radius), staleness, orphan pages, and broken links
• Searches via BM25 (FTS5), semantic embeddings, and hybrid RRF fusion

Install

npm install @wiki-nexus/core

Usage

import {
  WikiDatabase,
  indexWiki,
  computeClusters,
  traceKnowledgeFlows,
  computeImpact,
  searchHybrid,
  getPageContext,
  getStats,
} from "@wiki-nexus/core";

// Open or create a database
const db = WikiDatabase.open("wiki.db");

// Index a wiki directory
const result = await indexWiki(db, "/path/to/wiki", "my-wiki");
console.log(`Indexed ${result.totalPages} pages`);

// Run analysis
computeClusters(db);
traceKnowledgeFlows(db);
computeImpact(db);

// Search
const results = await searchHybrid(db, "knowledge management");

// Get full page context (360-degree view)
const context = getPageContext(db, "some-page-slug");

// Get wiki statistics
const stats = getStats(db);

db.close();

Wiki format

WikiNexus expects markdown wikis with:

• YAML frontmatter (title, tags, sources, updated)
• [[slug]] wikilinks where slug = filename without .md
• Flat page structure in wiki/pages/
• Optional SCHEMA.md at wiki root for detection

Part of WikiNexus

This is the core library. For the full CLI and MCP server, install wiki-nexus.

License

MIT