llm_wiki: Turn Documents Into an Interlinked Knowledge Base

Overview

LLM Wiki turns imported documents into a persistent knowledge base: an LLM creates and updates Markdown pages, links related ideas, and tracks their source material. This is an alternative to relying only on retrieval-augmented chat, where relevant passages are retrieved anew for each question.

It is aimed at people who want to build a reusable personal or project knowledge base from research papers, books, work documents, and web pages. The desktop app combines ingestion, search, chat, graph exploration, and human review, while keeping the resulting wiki compatible with Obsidian.

Key Features

• Ingests documents in formats including PDF, Office files, EPUB/MOBI, Markdown, images, and web clips.
• Uses a two-step LLM process to analyze sources, then generate wiki pages with source references and cross-links.
• Supports incremental ingestion, a persistent queue, and automatic monitoring of the source folder.
• Provides keyword search, graph-based expansion, and optional vector search.
• Offers chat with wiki and source retrieval, plus optional web research through configured search providers.
• Visualizes wiki relationships and clusters, and surfaces potential gaps and unexpected connections.
• Includes an asynchronous review queue so users can handle items flagged during ingestion.
• Exposes a local HTTP API and MCP server for compatible external tools and agents.

Use Cases

• Researchers can turn a growing collection of papers and reports into linked notes, then explore connections and gaps.
• Students and lifelong learners can maintain a searchable, evolving wiki from books, articles, and course material.
• Teams or professionals can organize project documents into a source-traceable knowledge base for later questions.
• Obsidian users can generate and browse Markdown wiki content while retaining access to the files as a vault.
• Developers can connect compatible agents to a locally running wiki through its API or MCP server.

Project Facts

• Language: TypeScript
• License: NOASSERTION
• Stars: 20.2k
• Forks: 2.3k
• Topics: none listed
• Archived: no

Getting Started

Download a platform build from the GitHub Releases, or build from source:

git clone https://github.com/nashsu/llm_wiki.git
cd llm_wiki
npm install
npm --prefix mcp-server ci && npm run mcp:build
npm run tauri dev

Building requires Node.js 20+, Rust 1.88+, and protoc. See the README for platform-specific setup, configuration, and usage details.

Alternatives

• AingDesk: AingDesk answers questions using knowledge bases and agents, while LLM Wiki turns documents into a persistent, linked wiki.

Considerations

• The app needs a configured LLM provider and model. API keys, model access, and any associated costs depend on the provider selected.
• Building from source requires both the Node.js and Rust toolchains, plus Protocol Buffers tooling.
• Vector search is optional and requires a configured embedding endpoint. Web research also requires setup for a supported search provider.
• Generated wiki content is LLM-produced. Source references, review tools, and source-only answering can help with verification, but users should check important claims against original documents.
• The repository lists its license as NOASSERTION, while the README says GPL-3.0. Confirm the repository's license files and terms before relying on or redistributing the software.