Run Local AI Models with Xinity and keinsaas Navigator

keinsaas Navigator now runs local AI models through Xinity. Alongside the major cloud models already available in Navigator, you can connect models running on your own server, your own workstation, or even your own PC, and power all your tools and automations from a single interface. No data leaves your infrastructure.

This article explains what the integration does, why local models belong in every serious AI strategy, and how to get started.

What is the keinsaas Navigator and Xinity integration?

keinsaas Navigator is a unified workspace for working with AI models and automating workflows. Like Xinity, it is open-source and can be self-hosted on your own server, so the entire stack, from the interface down to the inference engine, runs on infrastructure you control. Until now, Navigator connected to the major commercial model providers. With the Xinity integration, it also connects to models you host yourself.

Xinity is an open-source, on-premise inference engine with an OpenAI-compatible API. That compatibility is the whole trick: Navigator talks to your local Xinity endpoint exactly the way it talks to a cloud provider. Same interface, same workflows, different jurisdiction. Your prompts and documents are processed on hardware you control instead of being sent to a third party.

In practice, the setup looks like this: Xinity runs your chosen open-source model on your own machine, Navigator points at it, and every tool inside Navigator can use it immediately.

Why run AI models locally?

If local models are not part of your AI strategy, it is worth being honest about what your current setup actually is: a dependency.

Every workflow built purely on a commercial API depends on a single vendor. On their pricing, which can change. On their terms of use, which can change. On their availability in your region, which can change. And on their infrastructure, which is where your data ends up, often outside the legal space your customers and regulators expect it to stay in.

For companies in the EU, that last point is not abstract. GDPR obligations and the EU AI Act's transparency requirements both get simpler to reason about when inference happens on premises, because the logs, the data flows, and the audit trail are entirely yours.

Local models close that gap. And when both Navigator and Xinity run on your own servers, there is no part of the stack quietly phoning home. Cloud models remain useful, and Navigator keeps them available. The point is choice: route sensitive workloads to your own hardware and keep the rest where it is.

Open-source models caught up faster than expected

The reason this is practical now and was painful two years ago is the quality of open-weight models. With releases like Google's Gemma family, a large share of everyday business use cases, summarization, drafting, extraction, classification, internal Q&A, run well on models small enough for consumer hardware.

You no longer need a GPU cluster to make local AI useful. A capable workstation or a compact inference box is enough for many teams, and dedicated hardware like the ASUS Ascent GX10 takes it further for production workloads.

How to get started

Reach out to the keinsaas team to try the Xinity integration in Navigator and run local models alongside the cloud models you already use, or to get help setting it up inside your company.

If you want to go deeper on the infrastructure side, Xinity is open source under Apache 2.0 and runs anywhere from a single PC to dedicated on-premise hardware.

Frequently asked questions

What hardware do I need to run local models? For smaller open-source models like Gemma, a modern workstation or gaming PC is enough. For heavier production use, dedicated inference hardware is the better fit.

Does my data leave my infrastructure when I use a local model? No. With Xinity, inference happens on your own hardware. Prompts, documents, and outputs stay where you run them.

Can I self-host keinsaas Navigator? Yes. Navigator is open-source and can run on your own server, so both the interface and the Xinity inference engine stay within your infrastructure.

Is Xinity really open source? Yes, Apache 2.0 licensed. The code is public at github.com/xinity-ai/xinity-ai.

Which models can I run? Any open-weight model supported by the Xinity Runtime, including the Gemma family, Qwen, and other leading open-source models.