One Install Script for a Local AI Stack: Ollama, Open WebUI, n8n, ComfyUI
Setting up a local AI stack usually means stitching together four or five separate projects, each with its own install instructions, config files, and assumptions about how it should talk to everything else. Ollama wants one thing, Open WebUI wants another, and by the time you've got n8n and ComfyUI running you've spent a weekend on plumbing instead of actually using the tools. ODS (Osmantic Deployment System) is an attempt to collapse all of that into a single install script, and it's currently in public pre-release testing for V3.
What It Does
ODS installs and wires together the components you need to run AI locally, rather than making you assemble them by hand. That includes local model inference, a ChatGPT-style web UI you can reach from any browser, a control dashboard for managing models and services, voice and agent workflows, RAG and search over local documents, and local image generation. It also bundles privacy and ops tooling—service auth, secrets, observability, and diagnostics—into the same stack.
The repo is organized so the root holds the README, installers, security policy, and CI workflows, while the ods/ directory contains the actual product: services, installer phases, compose files, the dashboard, a CLI, tests, and operator docs. It runs on Linux, macOS, and Windows (via a guided Ubuntu/WSL2 setup). By default it goes online only to download models and container images, check GitHub for releases, and run web searches the Portal agent makes on your behalf. Inference, chat history, and your files stay local, and ODS collects no telemetry. Cloud and hybrid API modes exist but are optional.
Why It's Cool
It treats homelab setup as a solved problem. The README's framing is blunt about this: AI server setup is "rapidly becoming a solved problem" and should feel that way for everyone. That's a reasonable thesis. The individual pieces are mature—what's missing is the glue, and that's exactly what ODS is trying to be.
The privacy defaults are spelled out, not hand-waved. A lot of local AI projects claim "private by default" and leave it there. ODS lists the three specific cases where it goes online (model and image downloads, GitHub release checks, and Portal agent web searches) and links to an FAQ that explains how to turn each one off. That kind of specificity is refreshing.
One dashboard for the whole stack. Being able to manage models, services, setup, GPU status, and extensions from a single place is the practical payoff of bundling everything. If you've ever juggled four browser tabs and a terminal to figure out which service is misbehaving, this is the appeal.
It's honest about its own maturity. The README is upfront that the installers pull from main, which isn't a signed release and gets fixes continuously. There's a separate verified installer preview that stays gated until a signed release passes end-to-end testing. Known limits and validation status are documented in the release notes and validation docs rather than buried. For a pre-release, that's the right posture.
Update semantics are defined. Re-running the installer picks up code fixes; ods update refreshes container images only. Small detail, but it's the kind of thing that usually gets discovered the hard way.
How to Try It
The install is a single command per platform. Run it in a normal terminal.
Linux or macOS:
curl -fsSL https://install.osmantic.com/ods.sh | bash
Windows PowerShell uses a guided Ubuntu/WSL2 setup with Pixel/Portal. The README's Windows block is longer and includes a truncated section, so grab the full snippet from the repo rather than copying from a summary.
Once installed, ODS starts the stack, picks a model suited to your hardware, and gives you the local web UI. From there you can manage things through the dashboard.
A few things worth knowing before you run it:
- You're installing from
main, not a signed release. That means fixes land continuously, but you're on the bleeding edge. - Check the security advisories before pinning
v3.0.0, which is the latest published source release. - Read the release validation and installer trust docs if you want the full picture on what's been tested and what hasn't.
The repository is at github.com/osmantic/ods, and the README links to the release notes, FAQ, release channels, and installer trust documentation.
Final Thoughts
ODS is aiming at a real gap: the individual components of a local AI stack are solid, but assembling them is tedious and error-prone. If you've been putting off a local setup because the integration work looked like a chore, this is worth a look—just go in knowing it's a pre-release, not a signed, fully validated release. It's best suited to developers and homelab folks who are comfortable running install scripts and reading release notes before committing. The project's own framing is the right one: setup should feel solved. ODS is a serious attempt at making that true.
Follow @githubprojects for more developer tools and open source projects.