AutoRAG
Your Personal Librarian Agent. Ask anything on your computer, organize files, remember important info and files.
What is AutoRAG?
Your Personal Librarian Agent. Ask anything on your computer, organize files, remember important info and files.
Overview of AutoRAG
What it is
AutoRAG Agent is a personal librarian agent for office work. It searches files, mail, chats, and notes on your computer in place, opens source documents via bash to verify ground truth, and curates answers into numbered knowledge units. It is built on the Pi agent framework. The desktop app (Dazzi) provides a file manager interface with Quick Look previews and hybrid search. The legacy AutoRAG (RAG AutoML tool) is maintained in the legacy/ directory.
Key features
- Federated search across files, Slack, KakaoTalk, Discord, Telegram, WhatsApp, Mail, GitHub, and Obsidian vaults without migrating data.
- Hybrid keyword + meaning search with a reranker, and scheduled re-indexing that runs even while the app is closed.
- Duplicate file detection (Latest, Similar, Identical) powered by dupey.
- Quick Look preview for Markdown, code, PDF, video, audio, Office, and HWP/HWPX files (HWP rendered locally by rhwp).
- Tabs and split panes in the desktop app; folders update live as other apps change them.
- Private by design: semantic indexing runs on a local MinSync Qwen3-Embedding-0.6B engine; file contents are never sent to a remote embedder unless explicitly configured.
- Available surfaces: desktop app (Dazzi), terminal UI, messenger (Telegram, Discord), CLI, MCP server, TypeScript API, and agent skills.
Use cases
- Ask questions about your files, mail, chats, and notes and get answers with cited sources.
- Organize, rename, move, or edit files with a record of every change.
- Find duplicate or similar files across folders.
- Search across multiple data sources (Slack, KakaoTalk, Discord, Telegram, WhatsApp, Mail, GitHub, Obsidian) in one query.
Getting started
See the repository README for installation and quick-start instructions.
License and pricing
Licensed under Apache-2.0. Open source.
Based on the project's public README and repository metadata.