Nakyoung AI
Self-Hosted AI Chat Server with Web Search — 2026
End-to-End Developer (infrastructure, debugging, deployment) — self-initiated project
I wanted to run my own AI chat service — not a wrapper around someone else's API, but a model actually running on hardware I own and control. I turned my personal laptop into a server, running Gemma 4 and gpt-oss-20b through Ollama, accelerated on an NVIDIA RTX 4050, and exposed it publicly through Cloudflare Tunnel on a custom domain, without opening a single port.
On top of that, I built a custom-branded chat interface using Open WebUI, with real-time web search powered by a self-hosted search engine, SearXNG, aggregating results from Bing, Brave, and Wikipedia without a single paid API key. The whole service is orchestrated through a launcher script I wrote for starting, stopping, and checking status end to end.
Impact — Running in production today under my own domain, serving real conversations through a fully self-hosted stack — model, tunnel, search, and interface all owned end-to-end, with zero paid API keys.
