GD4AI/obsidian-llm-wiki

What it solves

This plugin solves the problem of disconnected, siloed notes in Obsidian. It automatically transforms a collection of scattered Markdown files, PDFs, and Office documents into a structured, queryable knowledge base (a "wiki") without requiring complex external infrastructure.

How it works

It implements a graph-based retrieval system inspired by the Karpathy LLM Wiki concept. Instead of relying on traditional vector databases or embedding models, it uses Personalized PageRank (PPR) algorithms to analyze the links between your notes. When you ingest files, the plugin uses an LLM to extract entities and concepts, creating new, interconnected wiki pages. You can then chat with this knowledge base via a side panel, with answers providing direct links back to your original notes.

Who it’s for

It is designed for Obsidian users who want to turn their "second brain" into an active, conversational knowledge base. It is ideal for those who prefer a local-first, self-contained tool that avoids the complexity of managing vector databases, Python environments, or separate server-side applications.

Highlights

• Zero-embedding architecture: Uses graph-based retrieval (PPR) instead of vector databases.
• Native Obsidian Integration: Runs entirely within Obsidian with support for the built-in Graph View, ribbons, and command palette.
• Multi-format Ingest: Supports Markdown, PDFs, and Office documents (DOCX, PPTX, XLSX) using built-in parsers like MinerU.
• Broad Model Support: Compatible with 16+ LLM providers, including OpenAI, Anthropic, Bedrock, Gemini, and local options like Ollama or LM Studio.
• Maintenance Tools: Includes automated health scans to fix duplicates, dead links, and orphan pages.