Query language

Thor query strings are used anywhere a GraphQL field accepts a query argument.

In practice, this powers search and filtering on connections like:

• products
• category.products
• collection.products
• admin search endpoints such as orders, customers, and products

The syntax is translated into Elasticsearch queries. It behaves closer to Shopify-style search than to SQL.

The base rules are:

• unprefixed terms search across indexed text fields
• field:value targets a specific field
• quoted values keep phrases together
• OR combines alternatives
• NOT or - excludes matches
• + requires a clause to match

A single token is treated like a prefix-capable search term. Multiple tokens are split and matched as separate prefix-friendly clauses.

Numeric and date fields support comparison syntax.

Thor rewrites inputs such as created_at:>='2025-10-01' into range queries before they reach Elasticsearch.

Common patterns:

• field:>=10
• field:<100
• created_at:>='2025-10-01'
• created_at:<'2025-10-15'

Query strings only filter what is already eligible for the request.

On the Storefront API, request context still decides the candidate set first, including store, publication channel, price channel, currency, country, and customer group. The query string then filters within that already-resolved catalog.