Query options

Control ranking, filtering, and compare behavior for Search().

interface GlyphQueryOptions {
  limit?: number;
  threshold?: number;
  normalize?: boolean;
  aggregate?: GroupResultAggregator;
  compare?: GlyphComparisonOptions;
}

`limit`

Maximum number of results returned after sort.

Value	Behavior
omitted	Return all results that pass `threshold`
`N`	Return top `N` hits

`threshold`

Minimum raw similarity (before normalize) to include a result.

Value	Default
omitted	`0` (include all non-negative scores)

Use a higher threshold to drop weak matches. Short queries against long documents often produce low absolute scores; threshold: 0 with normalize: true is common for ranking demos.

`normalize`

After sort (and limit), divide each similarity by the top score.

Value	Behavior
`false` (default)	Raw MinHash estimate
`true`	Best hit → `1.0`; others scaled proportionally

Skipped when:

No results remain, or
Top score is 0

comparison.similarity inside each result is updated to match.

`aggregate`

GroupResultAggregator used when an index entry (or the probe) is a GlyphGroup. Default: GroupResultAggregatorMax (max pairwise similarity).

import { index, query } from "@koda.oss/glyph";

const idx = index.New();
query.New(idx).Search(probe, {
  aggregate: ({ scores }) =>
    scores.reduce((a, b) => a + b, 0) / scores.length,
});

See Groups.

`compare`

Extra options forwarded to Compare() for each entry.

query.New(idx).Search(probe, {
  compare: {
    aggregate: GroupResultAggregatorMax,
  },
});

If both aggregate and compare.aggregate are set, the top-level aggregate wins.

Option interaction table

Order	Step
1	Compare each entry (apply `aggregate` / `compare`)
2	Filter by `threshold`
3	Sort descending
4	Apply `limit`
5	Apply `normalize` (if `true`)

Related links ranked by Glyph.