kdrant-core/dev.kdrant.model/PayloadIndexParams/Text

Text

@Serializable

@SerialName(value = "text")

data class Text(val tokenizer: Tokenizer? = null, val minTokenLen: Int? = null, val maxTokenLen: Int? = null, val lowercase: Boolean? = null, val phraseMatching: Boolean? = null, val asciiFolding: Boolean? = null, val stemmer: StemmingAlgorithm? = null, val onDisk: Boolean? = null, val memory: Memory? = null, val enableHnsw: Boolean? = null) : PayloadIndexParams

A full-text index, for matchText, matchTextAny and matchPhrase.

Constructors

Text

constructor(tokenizer: Tokenizer? = null, minTokenLen: Int? = null, maxTokenLen: Int? = null, lowercase: Boolean? = null, phraseMatching: Boolean? = null, asciiFolding: Boolean? = null, stemmer: StemmingAlgorithm? = null, onDisk: Boolean? = null, memory: Memory? = null, enableHnsw: Boolean? = null)

Properties

asciiFolding

@SerialName(value = "ascii_folding")

val asciiFolding: Boolean?

Fold accented characters to ASCII, so "ação" and "acao" are the same token.

enableHnsw

@SerialName(value = "enable_hnsw")

open override val enableHnsw: Boolean?

Whether Qdrant builds the extra HNSW links for this field, which is what makes a filtered search over it read one tenant's or one category's points instead of walking the collection and discarding. It needs payloadM above zero on the HNSW config to do anything, so a caller who sets one without the other has asked for half of a multi-tenant layout.

lowercase

@SerialName(value = "lowercase")

val lowercase: Boolean?

lowercase every token, making matching case-insensitive. Defaults to true.

maxTokenLen

@SerialName(value = "max_token_len")

val maxTokenLen: Int?

tokens longer than this are dropped.

memory

@SerialName(value = "memory")

open override val memory: Memory?

Memory placement of the index. Overrides onDisk when both are set.

minTokenLen

@SerialName(value = "min_token_len")

val minTokenLen: Int?

tokens shorter than this are dropped.

onDisk

@SerialName(value = "on_disk")

open override val onDisk: Boolean?

Whether Qdrant keeps this index on disk instead of in RAM. null leaves the server's default.

phraseMatching

@SerialName(value = "phrase_matching")

val phraseMatching: Boolean?

store token positions so matchPhrase can require them adjacent and in order. Defaults to false, and a matchPhrase filter against an index without it is accepted by the server and matches nothing.

stemmer

@SerialName(value = "stemmer")

val stemmer: StemmingAlgorithm?

How tokens are reduced to a stem. The server's default is no stemming.

tokenizer

@SerialName(value = "tokenizer")

val tokenizer: Tokenizer?

how a value is split into tokens; the server's default is Tokenizer.WORD.