#2626th of 51 in RAG, knowledge and search
autorag
Librarian agent that searches local files, mail and chats in place
- Stars
- 5.7k
- License
- Apache-2.0
- Last commit
- Nov 2026
- Last release
- Nov 2026
- Language
- TypeScript
4,678 on 21 Apr 20265,664 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
AutoRAG Agent (@autorag/librarian, TypeScript, Node 24+ or Bun) is a single-model agent built on the Pi framework. It searches local files, mail, Slack, Telegram, Discord and KakaoTalk without copying them into a central vector database. It combines SPLADE sparse, local vector and hybrid retrieval through MinSync with direct file reads via bash, and returns answers with numbered citations. It ships a CLI, TUI, MCP server and TypeScript API; the original Python RAG AutoML tool now lives in legacy/.
Who it is for: Engineers wanting cited answers over local files, mail and chats from CLI or MCP
Strengths
- Searches data in place across files, mail and chat apps, keeping source-native identities
- Local embeddings (Qwen3-Embedding-0.6B via MinSync); no remote embedder unless configured
- Surfaces: CLI, Pi-based TUI, MCP server, TypeScript API, Telegram/Discord/WhatsApp
- Parses HWP/HWPX, PDF, DOCX and XLSX in-process, no Java needed
Weaknesses
- Desktop app is a separate repo under Fair Source FSL-1.1-MIT, not Apache-2.0
- Needs a configured model provider or hosted plan; no bundled LLM
- Downloads MinSync and filee-cli (~85 MB) on first run
- SPLADE ranks exact identifiers among look-alike files less reliably than BM25
What it needs
- Needs Node.js >= 24 or Bun >= 1.2, MinSync, Pi coding agent, filee-cli, Jikji (optional, Rust), fsearch-cli (optional)
- Models: Provider-configured LLMs via Pi, Qwen3-Embedding-0.6B (local embeddings), KURE-Sparse-nano (SPLADE)
Quick answers
Is autorag open source?
Yes. Its code is on GitHub under the Apache-2.0 license.
Also in RAG, knowledge and search
See all 51| Rank | Project | Score |
|---|---|---|
| 1 | RAGFlowRAG engine with deep document parsing, template chunking and agentic retrieval | 90 out of 100 |
| 2 | LightRAGGraph-based RAG framework with a REST API server and web UI | 90 out of 100 |
| 3 | deeptutorSelf-hosted AI tutor with knowledge bases, guided learning, books and quizzes | 89 out of 100 |
| #4 | Open NotebookSelf-hosted NotebookLM alternative for research notebooks, chat, and podcasts | 89 out of 100 |
| #5 | MaxKBEnterprise knowledge-base agent platform with RAG, workflows and MCP tools | 87 out of 100 |