PrivateGPT

Fully offline

Air-gappable RAG over your docs. The OG offline-RAG project, now mature and team-friendly.

Editorial verdict: “Best when air-gap compliance is the requirement. Less polished than AnythingLLM, more configurable.

RAG app
Free
Apache-2.0
4.4 / 5
GitHub ★ 54,000

Compatibility at a glance

Which runtime + OS combos this app works against. Source of truth for "will it run on my setup?"

§ Runtimes supported
ollamallama-cppopenai-compat
§ OS / platform
linuxmacoswindows
§ Hardware + model hint
Minimum VRAM
8 GB
Recommended starter model
Llama 3.1 8B Q4_K_M + bge-small-en embedder

What it is

PrivateGPT is the right choice when your compliance posture demands a truly air-gapped RAG pipeline. It bridges to Ollama, llama.cpp, or any OpenAI-compatible endpoint, and runs on Linux, macOS, or Windows with at least 8 GB VRAM — a Llama 3.1 8B Q4_K_M paired with bge-small-en is the baseline. The ingestion pipeline is mature and the backend is swappable (LLM, embedder, vector store), but the UI is functional rather than polished; setup takes more effort than AnythingLLM. If you need reproducible offline retrieval over your docs and can trade interface gloss for configurable control, this is the proven foundation.

✓ Strengths

  • +Air-gap-friendly from day one
  • +Swappable LLM + embedder + vector store
  • +Active maintainer, large community

△ Caveats

  • UI is functional, not pretty
  • Setup is more involved than AnythingLLM