agents-features-longterm-memory/ai.koog.agents.longtermmemory.ingestion.extraction/FilteringExtractionStrategy

FilteringExtractionStrategy

class FilteringExtractionStrategy(val messageRolesToExtract: Set<Message.Role> = setOf(Message.Role.User, Message.Role.Assistant), val lastMessageOnly: Boolean = false) : ExtractionStrategy(source)

Default extractor that filters messages by role.

This extractor filters messages to only include those with roles in messageRolesToExtract, then converts each message's content into MemoryRecords.

When lastMessageOnly is true, only the last message for each role in messageRolesToExtract is extracted. This is useful with ai.koog.agents.longtermmemory.ingestion.IngestionTiming.ON_LLM_CALL to avoid re-ingesting messages that were already stored in previous calls.

Constructors

FilteringExtractionStrategy

constructor(messageRolesToExtract: Set<Message.Role> = setOf(Message.Role.User, Message.Role.Assistant), lastMessageOnly: Boolean = false)

Types

Builder for FilteringExtractionStrategy.

Properties

lastMessageOnly

val lastMessageOnly: Boolean

When true, only the last message for each matching role is extracted. Defaults to false.

messageRolesToExtract

val messageRolesToExtract: Set<Message.Role>

The set of message roles to extract and persist. Defaults to setOf(Message.Role.User, Message.Role.Assistant).

Functions

open suspend override fun extract(messages: List<Message>): List<TextDocument>

Transforms a list of messages into a list of memory records for storage.