AutoRAG

Your Personal Librarian Agent. Ask anything on your computer, organize files, remember important info and files.

Automatically checked · GitHub checked 2026-11-14

AutoRAG

Overview

AutoRAG Agent: your personal librarian agent. Dazzi is the desktop app.

AutoRAG Agent

Your personal librarian agent: it finds, reads, and cites anything on your computer.

npm version npm downloads GitHub stars CI Status Documentation License: Apache-2.0 Node >= 24 Bun >= 1.2

Dazzi: the AutoRAG desktop app

Dazzi is the app version of AutoRAG Agent. Download it for Mac and Windows at dazziapp.com, or build from NomaDamas/AutoRAG-app. It is a real file manager with the AutoRAG engine in this repository behind it.

Dazzi: a folder on the left, an AI Search conversation with cited answers on the right

  • Fast file search. Hybrid keyword + meaning search with a reranker finds files by what is inside them, and every answer links the passage it came from.
  • Messengers, mail, and more. Slack, KakaoTalk, Discord, Telegram, WhatsApp, every inbox you sign in to, GitHub, Obsidian, and your past coding-agent sessions (Claude Code, Codex, and 18 more) are searched alongside your files.
  • Preview anything. Press Space for Quick Look: Markdown, code, PDF, video, audio, Office, and HWP/HWPX, without opening another app.
  • File-based automations. Watch a folder, pick a trigger, and let the agent triage or organize new files for you.
  • Duplicate detection. final, final-v3, and (1) copies collapse into one row tagged Latest, Similar, or Identical.
  • Remote access. Open your running Dazzi from a phone or another computer over your Wi-Fi or an internet tunnel, paired with a 6-digit code or QR.
  • A real Finder. Tabs, split panes, light and dark mode, and Korean, English, Japanese, and Simplified Chinese.

[!IMPORTANT] Looking for the original AutoRAG (RAG AutoML / pipeline optimization tool)? This repository now hosts AutoRAG 2.0, a complete reimagining of AutoRAG as a personal librarian agent for your office work. The original Python-based AutoRAG — the RAG AutoML tool for automatically finding an optimal RAG pipeline for your data — now lives in the legacy/ directory of this repository.

The legacy AutoRAG is NOT abandoned. It continues to be maintained (bug fixes, dependency updates, and PyPI releases via pip install AutoRAG) in maintenance mode. Existing users can keep using it exactly as before — see the legacy README for its documentation, and file issues in this repository as usual. New feature development is focused on AutoRAG Agent (2.0).


What you can do in the app

🔎 Just ask

Ask the way you'd ask a teammate. The librarian reads the files and writes back, every claim linked to the passage behind it. Drag files, folders, or links into a question to ask about just those.

An answer about the Seoul office lease with numbered citations
A folder in list view with document versions stacked under one row

🗂️ Spot the twins

proposal-final, proposal-v3, proposal (1): stacked into one row, newest on top, tagged Latest, Similar, or Identical. Powered by dupey, the same scan as autorag duplicates.

⌨️ Space to see anything

Markdown, code, PDF, video, audio, Office, and HWP/HWPX previewed instantly in Quick Look. HWP pages are drawn locally by rhwp; Hancom Office is not required.

Quick Look showing a multi-page PDF with a page rail
Two folder panes side by side, each with its own tabs

🪟 A real file manager

Tabs and split panes: drag a tab to an edge to split, and every pane keeps its own tabs, so a folder and a conversation sit side by side. Folders update live as other apps change them.

And more: hybrid keyword + meaning search with a reranker · scheduled re-indexing that runs even while the app is closed · an Ask entry in the Finder and Explorer right-click menu · light and dark mode · Korean, English, Japanese, and Simplified Chinese.

Every place, one search

Supported sources: KakaoTalk, Slack, Discord, Telegram, WhatsApp, Mail, Notion, GitHub, Obsidian, browser history, RSS, Google Drive, iCloud Drive, Dropbox

Cloud drives (Google Drive, iCloud Drive, OneDrive, Dropbox, and others) are searched through their synced folders. Plus local documents, GitHub Gists, coding-agent sessions (via wombat: cargo install wombat-sessions, then add a wombat datasource), and photos. Data stays in each tool's own store. Full list and setup.

Private by design

  • Embedded on your machine. Semantic indexing runs on MinSync's local Qwen3-Embedding-0.6B engine (1024 dimensions), which MinSync provisions and runs; AutoRAG owns no embedding model. File contents are never sent to a remote embedder unless you explicitly configure a remote minSync.embedder.
  • Keys stay encrypted. Sign-in tokens and API keys live in the system's encrypted store, never in a config file.
  • Bring your own model. Use the hosted Dazzi Models plan, or connect your own provider key and pay the provider directly.

Install it with your agent

Skip the manual steps: paste a prompt into your coding agent (Claude Code, Codex, Cursor, Pi, ...). The agent reads the matching skill and does the setup, asking before it touches anything.

Set up the AutoRAG librarian (autorag-setup):

Install and configure AutoRAG Agent on this computer by following
https://raw.githubusercontent.com/Marker-Inc-Korea/AutoRAG/main/skills/autorag-setup/SKILL.md
Ask me which folders to index and how to sign in to a model before you change anything.
When it is set up, run a health check and a test search, and show me the result.
Never move, rename, edit, or delete my documents while setting up.

Give your coding agent model-free search over your files (autorag-lite-setup):

Install and register the AutoRAG Lite MCP server for this agent by following
https://raw.githubusercontent.com/Marker-Inc-Korea/AutoRAG/main/skills/autorag-lite-setup/SKILL.md
Ask me which folders to index, build the index, then verify with a real search.
Never move, rename, edit, or delete my documents.

Something broken? (autorag-doctor):

Check AutoRAG by following https://raw.githubusercontent.com/Marker-Inc-Korea/AutoRAG/main/skills/autorag-doctor/SKILL.md
and finish with the per-datasource status table.

Prefer to do it yourself? See Installation and Agent setup.

One engine, every surface

The app is one way in. The same librarian runs everywhere else:

SurfaceUse it forStart here
Desktop app (Dazzi)Browsing and asking in one windowNomaDamas/AutoRAG-app (Fair Source, FSL-1.1-MIT)
Terminal UIInteractive, steerable search sessionsautorag tui
MessengerAsking the librarian from a chat appTelegram, Discord
CLIScripts, indexing, health checksCLI Command Reference
MCP serverFeeding retrieval to Claude Code, Codex, and other agentsAutoRAG Lite MCP Server
TypeScript APIEmbedding the librarian in your own appProgrammatic usage
Agent skillsLetting a coding agent install and drive AutoRAGFor AI Agents

More

Community

Built on Pi, MinSync, Jikji, dupey, kordoc, and more; see licenses and acknowledgements.

License

AutoRAG Agent is Apache-2.0; legacy legacy/ is Apache-2.0. Third-party notices: NOTICE, details.