Skip to content

Home

Jump to bottom
rzfz.ai release bot edited this page Jul 29, 2026 · 3 revisions

The rzfz.ai Stack — Community Wiki

The rzfz.ai Stack — the integrated open-source stack for local AI infrastructure. It brings a chat UI, private LLM inference, workflow automation, single sign-on, and monitoring together into one self-hosted platform that runs entirely on hardware you control.

This is the Community Wiki (on GitHub) — the public documentation for the Community stack. It covers getting started, the architecture, and installing the stack yourself. It is free to read, and the projects it documents are free to run for private and evaluation use.

📢 What's new & release notes

The stack ships on a monthly General Availability cycle, with patch releases in between. Start here to see what changed:

These are the same notes shown in your box's in-product Configuration Portal.

What it is

The rzfz.ai Stack is a self-hosted, modular, Docker-Compose AI service stack. Everything runs as containers on a single Linux host, and you turn capabilities on and off at the module level via Docker Compose profiles. Out of the box you get:

  • Chat UI — Open WebUI for talking to your models.
  • LLM inference — GPUStack serving language models locally (AMD, NVIDIA, or CPU).
  • Workflow automation — Dify for building LLM apps and pipelines.
  • Identity & single sign-on — Authentik in front of every app.
  • Monitoring, search, document tools, agents, and more — around 25 optional modules, each a Compose profile you enable when you need it.

Two design choices define the stack:

  • Caddy is the single entry point. One process listens on ports 80/443, terminates TLS, and gates every app.
  • Authentik provides single sign-on. You log in once, and Authentik decides which apps you can reach.

See the Architecture overview for how these fit together.

Who it's for

The stack is for teams and individuals who want modern AI capabilities — chat, retrieval, workflows, agents — without sending their data to a third-party cloud. Everything stays on your own appliance or server. It suits privacy- conscious organisations, on-premises deployments, evaluation labs, and anyone who prefers to own their AI infrastructure.

Open-core, in one paragraph

The stack follows an open-core model. The Community components are Apache-2.0 — genuine open source, free for any use, including private and evaluation use with no key and no registration. The first-party enterprise and orchestration components are source-available and their commercial production use is governed by the rzfz.ai Subscription (which, on a rolling Change Date, converts each released version to Apache-2.0). The many bundled upstream projects (Open WebUI, Dify, Authentik, GPUStack, and others) each keep their own upstream licences. In short: read, run privately, and evaluate freely; a Subscription covers commercial production of the source-available parts.

Community documentation

  • Getting started — prerequisites at a glance and the three-line quickstart.
  • Architecture overview — the module system, the always-on core, networking, and how a request flows.
  • Community install — install the Community stack yourself, from cloning the code to opening the UIs.

Where to go next

  • Enterprise documentation → docs.rzfz.ai — the gated operations manual (operations, upgrade runbooks, hardening, SSO and compliance). It is part of the rzfz.ai Subscription and is not part of this public wiki.
  • Community support → GitHub Issues — report a bug or ask a question, publicly and free.
  • Source code → GitHub — the public source repository for the Community stack.

Clone this wiki locally