VectifyAI/OpenKB

What it solves

OpenKB transforms raw, fragmented documents into a structured, interlinked wiki-style knowledge base. Unlike traditional RAG, which retrieves information from scratch for every query, OpenKB compiles knowledge once into a persistent, evolving wiki where information compounds and cross-references are maintained automatically.

How it works

OpenKB uses LLMs to process documents (PDFs, Word, Markdown, URLs, etc.) and compile them into summaries, concept pages, and entity pages (people, organizations, places, products) using plain Markdown files with wikilinks.

For long documents (typically 20+ pages), it integrates with PageIndex to create a hierarchical tree index, allowing the LLM to reason over the document structure rather than reading the full text. This vectorless, reasoning-based retrieval ensures accuracy for complex files. The resulting wiki is Obsidian-compatible and follows the Google Open Knowledge Format (OKF) specification.

Who it’s for

It is designed for users who need to manage large volumes of technical documents, research papers, or corporate knowledge, and want a persistent, human-readable knowledge base that can be integrated with other AI agents (like Claude Code or Gemini CLI) as a redistributable "skill."

Highlights

• Vectorless Retrieval: Uses PageIndex tree indexing for long documents instead of a traditional vector database.
• Compiled Wiki: Automatically generates and updates interlinked Markdown pages for concepts and entities.
• Multi-modal Support: Capable of retrieving and understanding figures, tables, and images.
• Skill Factory: Distills knowledge from the wiki into portable agent skills for other LLM-based tools.
• Obsidian Integration: The wiki is stored as plain .md files, making it natively compatible with Obsidian's graph view.
• Knowledge Workbench: Includes a bundled web UI for browsing, uploading, and querying the knowledge base.