Lumina-Match

A small TypeScript/JavaScript library for fast user-input suggestions using a two-stage pipeline:

1. N-gram filter (bigrams) to quickly reduce candidates
2. Fuzzy ranking (prefix/token overlap/Levenshtein) to sort suggestions

Install

npm install lumina-match

Usage

import { createMatcher } from 'lumina-match';

const myDataSet = [
  { id: 'A-12', phrase: 'Wireless Keyboard Layout' },
  { id: 'B-44', phrase: 'Mechanical Switches Blue' },
  { id: 'C-01', phrase: 'USB-C Charging Cable' }
];

const matcher = createMatcher(myDataSet);

console.log(matcher.suggestions('Kybord'));
// => [{ id: 'A-12', phrase: 'Wireless Keyboard Layout', confidence: 78 }]

More examples

Get more / fewer suggestions

import { createMatcher } from 'lumina-match';

const matcher = createMatcher(myDataSet);

matcher.suggestions('usb', 3); // top = 3
matcher.suggestions('usb', 20); // top = 20

One-off helper (no matcher reuse)

import { suggestions } from 'lumina-match';

suggestions(myDataSet, 'mech swtch', { top: 5 });

Use EntityMatcher directly

import { EntityMatcher } from 'lumina-match';

const engine = new EntityMatcher(myDataSet);
engine.suggest('Kybord', 10); // limit = 10

Typical UI pattern (reuse matcher)

import { createMatcher } from 'lumina-match';

const matcher = createMatcher(myDataSet);

export function onUserInput(value: string) {
  // Library returns [] for empty/very short queries.
  return matcher.suggestions(value, 10);
}

API

• createMatcher(data).suggestions(input, top = 10)
• new EntityMatcher(data).suggest(query, limit = 5)
• suggestions(data, input, { top })

Data types

• DataItem: { id: string; phrase: string }
• Suggestion: { id: string; phrase: string; confidence: number } (0–100)