Parity Features

Fuse.js Compatibility

Normalized Scores

Enable includeScore: true to receive 0-1 normalized scores instead of raw additive values:

const results = searchSections(index, {
  query: 'search',
  includeScore: true
});
// results[0].score: 0.95 (normalized between 0 and 1)

Match Spans

Use includeMatches: true to get character ranges for highlighting:

const results = searchSections(index, {
  query: 'install',
  includeMatches: true
});
// results[0].matches: [{ key: 'headingText', start: 0, end: 7 }]

Extended Query Syntax

Search supports advanced operators:

// Exact phrase match
searchSections(index, { query: '"exact phrase"' });

// Exclude terms (NOT or -)
searchSections(index, { query: 'install NOT beta' });
searchSections(index, { query: 'install -beta' });

// OR operator
searchSections(index, { query: 'install OR setup' });

Field Weights

Custom scoring weights via weights option:

const results = searchSections(index, {
  query: 'search',
  weights: { headingText: 100, bodyText: 50 }
});

BM25 Scoring

Enable BM25 (TF-IDF) scoring for more principled relevance ranking across large indexes:

const results = searchSections(index, {
  query: 'installation',
  scoringAlgorithm: 'bm25',
  includeScore: true
});

BM25 considers:

• Document frequency of query terms
• Field length normalization (shorter, more focused matches rank higher)
• Term frequency with saturation

Autocomplete / Suggest

Get search suggestions as users type:

const suggestions = suggest(index, 'inst');
// ['installation', 'instructions', 'instance']

This powers inline suggestions in custom implementations.

Faceted Search

Get counts of indexed records by type:

const counts = facets(index);
// { section: 150, action: 24, field: 12, link: 8 }

Use with filter to build faceted UIs:

const results = searchSections(index, {
  query: 'search',
  filter: (r) => r.type === 'action'
});

Query Analytics

Track popular queries locally (no backend required):

// Track when users perform searches
trackQuery(index, currentQuery);

// Get most popular queries
const popular = getPopularQueries(index, 10);
// [{ query: 'install', count: 42 }, ...]

Useful for analytics dashboards and improving search UX.

Index Manipulation

Index Mutations

Remove or update records after indexing:

// Remove a record
removeFromIndex(index, 'record-id');

// Update/replace a record
updateRecord(index, {
  id: 'record-id',
  url: '/new-url',
  headingText: 'New Title',
  bodyText: 'New content...',
  type: 'section'
});

Serialization

Export and import indexes for pre-built search:

// Serialize index to JSON
const json = serializeIndex(index);

// Restore from JSON
const restored = deserializeIndex(json);

Prebuilt Index

Load a pre-built index instead of crawling:

// In browser config
const config: ReefConfig = {
  prebuiltIndexUrl: '/search-index.json'
};

Text Processing Pipeline

Stemming & Stop Words

Enable tokenization pipeline for stemming, stop words, and diacritic folding:

const config: ReefConfig = {
  tokenizePipeline: [
    (tokens) => tokens.map(t => t.toLowerCase()),
    (tokens) => tokens.filter(t => !stopWords.has(t)),
    (tokens) => tokens.map(t => stem(t))
  ]
};

Synonym Expansion

Map alternate terms to improve recall:

const config: ReefConfig = {
  synonyms: {
    'log out': ['sign out', 'logout'],
    'log in': ['sign in', 'login'],
    'delete': ['remove']
  }
};

Diacritic Folding

Automatic accent normalization (café ↔ cafe) can be enabled in the tokenization pipeline.

Feature Comparison