OneSearch is built around a simple idea: search should return a structure, not only a ranked list. It clusters fetched results locally and labels groups with Ollama, while keeping every source page one click away.
Before the July 28 launch, I want to understand where this workflow would be most valuable: research, shopping, technical troubleshooting, or something else?
I'd also value thoughts on the trade-off between a lightweight 4B default model and optional 8B/14B models. What would make you trust an AI-created folder label enough to use it?
OneSearch turns flat web results into semantic folders using local AI. It fetches sources, creates local embeddings, clusters related pages, and labels each group with Ollama. The desktop app keeps the folder structure visible while sources open in separate tabs. Free users can run a lightweight 4B model; Pro unlocks 8B or 14B choices. Available for macOS Apple Silicon and Windows x64.
