AutoRAG
Developed by Marker-Inc-Korea
AutoRAG Agent is a personal librarian agent from Marker-Inc-Korea built on the Pi agent framework. A single configured model searches your files, mail, chats, and notes in place, opens source documents to verify ground truth, and curates answers into numbered, cited knowledge units. It federates data in place via CLI-native stores (Slack, KakaoTalk, Discord, Telegram, WhatsApp, GitHub, Obsidian) instead of migrating to a vector DB. Retrieval combines SPLADE sparse, dense vector, and hybrid ranking with MinSync incremental CDC chunking and local Qwen3 embeddings—private by design with zero telemetry. It can also organize, rename, move, or edit files on request with a full change record.
- In-place federated retrieval, no data migration
- SPLADE sparse + vector + hybrid multi-recall
- Local embeddings with incremental CDC indexing
- Numbered citations with source verification
- Acts on files: organize, rename, move, edit
desktopweb