llm-wiki-agent: what it is, what problem it solves & why it's gaining traction

llm-wiki-agent: what it is, what problem it solves & why it's gaining traction

What it solves

LLM Wiki Agent is a coding agent skill that transforms a collection of raw documents into a structured, interlinked wiki. Unlike traditional RAG (Retrieval-Augmented Generation) systems that re-derive knowledge on every query, this tool compiles knowledge into persistent markdown pages that accumulate and compound over time, automatically handling cross-references and flagging contradictions between sources.

How it works

The system operates as a set of instructions (schema files like CLAUDE.md or AGENTS.md) that guide a coding agent (such as Claude Code, Codex, or Gemini CLI) to manage a local directory of markdown files.

  • Ingestion: It reads source documents (converting non-markdown files like PDFs, DOCX, and PPTX via markitdown) and extracts knowledge to create entity pages (people, companies), concept pages (ideas, frameworks), and a living overview.
  • Synthesis: It can answer queries by synthesizing information across the wiki and optionally saving those answers as new synthesis pages.
  • Knowledge Graph: It generates a graph.html visualization using NetworkX and vis.js, mapping both explicit [[wikilinks]] and inferred semantic relationships.
  • Maintenance: A "lint" command identifies orphan pages, broken links, and data gaps.

Who it’s for

  • Researchers: To build a structured reference of papers and reports over several weeks.
  • Readers: To track characters, themes, and arguments in books.
  • Knowledge Workers: To maintain a personal knowledge base or business intelligence from meeting transcripts and project docs.
  • Analysts: To perform competitive analysis by tracking companies and technologies over time.

Highlights

  • Persistent Knowledge: Knowledge is stored in plain markdown, making it compatible with tools like Obsidian.
  • Automatic Structuring: Auto-creates entity and concept pages without manual writing.
  • Capstone Synthesis: Flags contradictions between sources at the time of ingestion.
  • Multi-format Support: Ingests a wide variety of formats including PDF, DOCX, XLSX, and EPUB.
  • Visual Knowledge Graph: Provides an interactive HTML visualization with community detection to cluster related topics.

Sources