String types

Keyword

Keyword treats a string as a single, indivisible value — the same way string fields work in relational databases. It supports term query, terms query, range query, wildcard query, prefix query, and exists query (null value query), as well as sorting and statistical aggregation (including GroupBy).

Good candidates for Keyword: user IDs, status codes, email addresses, country codes, and any field where you query exact values or run aggregations.

On datasets with more than one million entries, wildcard and prefix query performance degrades as the dataset grows. Use FuzzyKeyword if high-performance fuzzy matching at scale is required.

Text

Text is designed for full-text search. Before indexing, the system splits a string into tokens using the tokenization method you configure, then indexes each token individually. At query time, the query string goes through the same tokenization before matching against the index.

Because Text indexes tokens rather than the original value, it supports relevance scoring, result highlighting and summary, match query, and match phrase query. It does not support exact term queries, sorting, or aggregation.

Good candidates for Text: product descriptions, article content, user reviews, and any field where you search for words or phrases within a longer string.

For available tokenization methods, see Tokenization.

FuzzyKeyword

FuzzyKeyword is optimized for wildcard, prefix, and suffix queries. It pre-stores substrings of each field value in the index, so queries scan a compact index structure rather than scanning values row by row. Wildcard and prefix query performance stays consistently fast regardless of dataset size.

Good candidates for FuzzyKeyword: log messages, file paths, product SKUs, and any field where users type partial strings to search.

For suffix queries, use SuffixQuery with the FuzzyKeyword type. An alternative — reversing the data and running PrefixQuery on a Keyword field — is available but delivers lower performance.