A Man Built Private AI Search On A Mini PC. The Architecture Is Trivial. The Discipline Is Not.
A How-To Geek author combined Open WebUI, Ollama, and Syncthing on a Windows 11 mini PC with an AMD Ryzen 5 7640HS processor, 32GB of RAM, and Radeon 760M integrated graphics. Open WebUI provides the search interface. Ollama runs a local model. Syncthing keeps a shared folder synchronized across devices without uploading anything to a third-party cloud. The result is a private AI document search engine accessible from any PC in the house.
This demonstrates what I call local-first architecture. The mechanism: Open WebUI can only search documents it can see, so keeping the shared folder current is as critical as the model itself. Syncthing handles synchronization without a cloud intermediary, which means your data never leaves your hardware. The lesson for the layperson is that privacy is not a feature you buy. It is a system you architect by choosing tools that respect your boundaries.
The author runs this setup on a Windows 11 machine with an AMD Ryzen 5 7640HS, 32GB of RAM, and Radeon 760M integrated graphics. Not a dedicated AI workstation. Sufficient nonetheless. Syncthing, Open WebUI, and Ollama are all free, which removes the cost excuse entirely.
- Download and install Ollama from ollama.com on any PC with at least 16GB of RAM. Pull a small model like llama3 or mistral using the command line. This gives you a local AI model running entirely on your machine.
- Install Open WebUI, which connects to Ollama and gives you a chat interface. Upload a few personal documents into its knowledge base and ask it questions about them. You now have local AI document search on one machine.
- Install Syncthing on your main PC and on a second device. Share a folder between them. Place your documents in that shared folder so both machines always have the latest copies without using any cloud service. You have now replicated the architecture, if not the full multi-device search setup, in under 15 minutes.