Skip to main content

Semantic Search

Semantic search helps Qluent understand questions that don't exactly match your data values.

Your database might have country = 'United States of America', but users might ask about "US", "USA", or "American customers". Semantic search bridges this gap using AI to match similar meanings.

Use case

Enable semantic search on columns where users might use different words than what's in the data:

User saysDatabase has
"Nike shoes"brand = 'Nike, Inc.'
"completed orders"status = 'delivered'
"California"state = 'CA'

How it works

  1. You enable semantic search on a column
  2. Qluent extracts all distinct values from that column
  3. Each value gets an AI embedding (a numeric representation of its meaning)
  4. When users ask questions, Qluent finds the closest matching values

This happens automatically in the background after you enable semantic search.

Enable semantic search on a column in the data model settings.

Good candidates

  • Geographic columns - country, region, city, state
  • Product/category names - product_name, category, brand
  • Status/type columns - order_status, customer_type
  • Names users reference loosely - campaign_name, channel

Columns to avoid

  • High-cardinality columns - customer_id, transaction_id (too many values)
  • Numeric columns - price, quantity, age
  • Date columns - created_at, order_date
  • Boolean columns - is_active, has_subscription

Tips

  • Start with key categorical columns (country, product, category)
  • Combine with good descriptions—semantic search helps match values, descriptions help Qluent choose the right column
  • Processing takes time for columns with many values