Skip to main content
Learn how to optimize search performance for your use case.

​
Use Appropriate topK

Request only as many results as you need:
src/performance-topk.ts
// Good - only fetch what you'll display
const results = await toolIndex.search(query, { topK: 5 });

// Wasteful - fetching more than needed
const results = await toolIndex.search(query, { topK: 1000 });

​
Use Filters to Reduce Search Space

Apply metadata filters to narrow results:
src/performance-filters.ts
// Filter first, then rank - more efficient mental model
const results = await toolIndex.search(query, {
  filter: (m) => m.owner === 'billing',
  topK: 10,
});

​
Enable HNSW for Large Datasets

For datasets > 10,000 documents, enable HNSW indexing:
src/performance-hnsw.ts
const db = new VectoriaDB({
  useHNSW: true,
  hnsw: { efSearch: 50 },
});

​
Performance Comparison

DocumentsBrute-forceHNSW (ef=50)
10,000~50ms~1ms
50,000~250ms~1ms
100,000~500ms~2ms
See HNSW Scaling for details.

​
Query Caching

Cache frequent queries for better performance:
src/query-caching.ts
const queryCache = new Map<string, SearchResult[]>();

async function cachedSearch(query: string) {
  const cacheKey = query.toLowerCase().trim();

  if (queryCache.has(cacheKey)) {
    return queryCache.get(cacheKey)!;
  }

  const results = await db.search(query);
  queryCache.set(cacheKey, results);

  // Clear cache after 5 minutes
  setTimeout(() => queryCache.delete(cacheKey), 5 * 60 * 1000);

  return results;
}

​
Warmup Queries

Run a warmup query during startup to optimize the embedding pipeline:
src/warmup.ts
async function warmupSearch() {
  await db.initialize();

  // Warmup embedding pipeline
  await db.search('warmup query', { topK: 1 });

  console.log('Search ready');
}

​
Batch Similar Queries

If you need to run multiple searches with the same query:
src/batch-queries.ts
// Instead of running multiple searches
const allResults = await db.search(query, { topK: 100 });

// Then filter in memory
const billingResults = allResults.filter(r => r.metadata.owner === 'billing');
const userResults = allResults.filter(r => r.metadata.owner === 'users');

​
Monitoring Search Performance

src/performance-monitoring.ts
async function monitoredSearch(query: string, options?: SearchOptions) {
  const start = Date.now();

  const results = await db.search(query, options);

  const latency = Date.now() - start;

  console.log({
    query: query.substring(0, 50),
    latencyMs: latency,
    resultCount: results.length,
    topScore: results[0]?.score,
  });

  return results;
}

​
Performance Checklist

1

Use appropriate topK

Don’t request more results than you need
2

Apply metadata filters

Reduce the search space with filters
3

Enable HNSW for scale

Use HNSW indexing for > 10,000 documents
4

Cache frequent queries

Implement query caching for repeated searches
5

Warmup on startup

Run a warmup query to optimize the pipeline

HNSW Overview

Scale with HNSW

HNSW Tuning

Tune HNSW parameters

Basic Search

Search fundamentals