GitHub - joelhooks/pdf-library: Local PDF knowledge base with vector search using PGlite + pgvector
Local PDF knowledge base with vector search using PGlite + pgvector - joelhooks/pdf-library
CLI Search Tool
Summary
QMD is a command‑line tool that indexes markdown notes, meeting transcripts and documentation for local search. It supports BM25 keyword search, vector semantic search, and LLM re‑ranking, and can be called directly or via an MCP server for agentic workflows.
Highlights
- Combines fast BM25, embeddings‑based semantic retrieval, and optional LLM reranking
- All processing runs locally using node‑llama-cpp with GGUF models
- Simple commands to add collections, generate embeddings, search, and export results
- Provides JSON and file list outputs designed for integration with AI agents
- Offers an MCP server for programmatic access and tighter tool‑agent coupling
auto-generated
Context
Audience
Developers and power users who manage personal knowledge bases and need on‑device, privacy‑preserving search integrated with AI workflows
DomainLocal knowledge base search and retrieval
Formatcommand-line application with optional MCP server
Accessopen source on GitHub, installable via npm or bun
Topics
Related
ObsidianRoam ResearchLlamaIndexsemantic‑search libraries for Node.jsLLM agents frameworks