Our mission is to actively support the development and practical adoption of local AI, so that more people can use AI without giving up control of their data to external cloud services.
Cloud AI is fine for public material. But contracts, finances, client data, and private notes deserve a tool that works for you without reporting about you.
Open models already run well on everyday hardware. What's missing is practical guidance and tools that don't require a weekend of tinkering. That's the gap we work on.
Privacy claims should be verifiable, not marketing. That's why the software is open source (GPLv3), and why this site itself runs without analytics or advertising services and without third-party scripts (a single strictly necessary security cookie from the host aside; see the privacy policy).
Clear guides on when local AI is worth it, which hardware is enough, and which models fit which tasks, growing into a practical knowledge hub.
The HilbertRaum Software: a free, open-source local AI workbench for documents and private knowledge.
The HilbertRaum AI Kit: the same stack, preconfigured on a plug-and-play kit: local AI without setup friction.
The product is close to launch. Right now we're collecting the waitlist for the first batch of kits. Early users directly shape model selection, default workflows, and the first guides.
A Hilbert space (German: HilbertRaum, after the mathematician David Hilbert) is a space that is complete: every convergent process arrives inside the space, never outside it. We couldn't find a better description of how AI should treat your documents. And for the mathematically inclined: the document embeddings our software searches live in exactly such spaces. The name isn't a metaphor. It's a specification.
Nothing leaves the space.