I mean, I could write one! I kind of just pieced it together from guides on the three individuals

Edit: back of the napkin guide below is basically in the OpenWebUI docs already! I use NixOS (btw) but docker/podman should work well.

OpenWebUI + Ollama setup – tl;dr docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main

OpenWebUI SearXNG guide – a little more involved, but not difficult.

You can self host that too ;)

OpenWebUI + Ollama + SearxNG. OpenWebUI can do llm web search using the engine of your choice (even self hosted SearxNG!). From there it’s easy to set the default prompt to always give you the top (10, 20, whatever) raw results so you’re not confined to ai results. It’s not quite duck.ai slick but I think I can get there with some more tinkering.

My understanding is it scrapes what it can’t meaningfully get out of an API. Public instances run into rate limiting, but private instances don’t really have that problem.