documentcloud/visualsearch — reverse-engineered prompt

Build me a browser based rich search box component that makes it easy to create faceted searches from real data. I want a search input where people can type things like author, date, tag, status, and get autocomplete for both the facet names and the allowed values. It should feel smooth and visual, with each chosen filter turning into a clear little token in the search box, and it should support adding, editing, and removing filters without fuss.

The important part is that the final query can be read back as a clean structured object, so an app does not have to parse a messy text string. Please include a simple demo page that shows a few sample facets and values, plus basic docs or examples showing how someone would configure the available facets and hook into the query output.

If anything is unclear, look up the current docs or examples online and recreate the intended behavior in a polished way.