How the demo calculates article word count, builds numeric buckets, and turns them into live sidebar filters.

Article Length Facets in the Demo

The demo now exposes article length as a numeric facet in the right sidebar.

That adds a different kind of navigation signal than tags or sections:

How the value is created

The build step strips markdown to plain text and counts words per document.

That value is stored as wordCount in the generated demo payload and indexed with NumericFieldIndex.

Conceptually, the build pipeline does this:

const body = stripMarkdown(markdownBody);
const wordCount = body.split(/\s+/).filter(Boolean).length;

Then the numeric field is indexed like any other structured field:

fields: {
  wordCount: [String(entry.wordCount)]
}

At runtime, the sidebar computes numeric aggregations over the currently selected result ids.

The demo uses:

The range buckets are explicit:

[
  { key: "short", to: 400 },
  { key: "medium", from: 400, to: 800 },
  { key: "long", from: 800, to: 1400 },
  { key: "deep", from: 1400 }
]

This keeps the labels stable even as the result set changes.

When a user clicks an article-length chip, the demo does not run a special side query.

It updates the normal search state and adds a RangeQuery("wordCount", ...) filter.

That means:

So the sidebar stays consistent with the visible results.

Length is not a topical facet, but it is often a useful navigation hint in docs: