gknowledge

•
29
•

Transform documents into an interconnected, AI-maintained knowledge base with native GNOME integration and offline-first capabilities.

gknowledge brings Andrej Karpathy's LLM Wiki pattern to the Linux desktop as a native GTK 4 and Libadwaita application. Designed around local-first principles, it automatically ingests documents to build and maintain cross-linked Markdown wikis compatible with Obsidian. The application operates completely offline using a built-in heuristic extraction engine without requiring an API key, while also supporting external LLMs such as Ollama, local OpenAI-compatible endpoints, or ChatGPT accounts. By providing a native GNOME interface and a built-in Model Context Protocol (MCP) server, gknowledge allows users and external AI coding agents to search, query, and grow structured knowledge graphs with provenance and citations.

Key Features

01Native GTK 4 and Libadwaita desktop interface following GNOME Human Interface Guidelines
02Offline-first design with a built-in heuristic extraction engine requiring no API keys or accounts
03Obsidian-compatible plain Markdown storage with YAML frontmatter, wikilinks, and SQLite indexing
04Bring-your-own-model support including Ollama, local endpoints, OpenAI-compatible APIs, and ChatGPT
05Built-in MCP server enabling external agents like Claude Code, Cursor, and Codex to access and edit the wiki
0629 GitHub stars

Use Cases

01Turn collections of research papers, documentation, or articles into linked, auto-synthesized reference vaults
02Expose local Markdown knowledge bases to AI agents via MCP for context-aware query answering and research updates
03Maintain a fully private, offline 'second brain' without sending proprietary data to cloud providers