I really needed a RAG setup for work docs that can't leave our servers – no OpenAI, no cloud leaks, nothing. So I ended up building this: RAG-Enterprise.
It's all local, zero external calls.
What it does (the parts that seem solid so far):
Setup is dead simple: ./setup.sh standard – grabs Docker, Ollama, Qdrant, FastAPI + React frontend. Takes maybe 15-60 mins depending on your machine and connection
Users log in with JWT, there are roles (basic user, power user, admin)
Handles PDF with real OCR (Tesseract does the heavy lifting), plus Word, PPT, Excel, markdown... Apache Tika parses most stuff
Backups/restores via rclone to tons of places – S3, Google Drive, OneDrive, SFTP, whatever (like 70 options)
Inference runs on Ollama – whatever models you pull (Llama 3.1, 3.2, Mistral, Phi...)
LangChain for chunking/embedding/retrieval, Qdrant as vector store
Clean chat UI to query your uploaded documents
License is AGPL-3.0. I get it, some orgs see that and nope out – if it gets traction maybe dual-license down the line.
primoco•1h ago
I really needed a RAG setup for work docs that can't leave our servers – no OpenAI, no cloud leaks, nothing. So I ended up building this: RAG-Enterprise.
It's all local, zero external calls.
What it does (the parts that seem solid so far):
License is AGPL-3.0. I get it, some orgs see that and nope out – if it gets traction maybe dual-license down the line.Repo, docs and a quick GIF in the README: https://github.com/I3K-IT/RAG-Enterprise
Would love real feedback, the honest/rude kind especially:
Thanks a ton for checking it out – hit me with questions or fixes!