actrone-memory
    Preparing search index...

    Interface L2SearchParams

    A ranked search hit: the entry plus its blended relevance score.

    interface L2SearchParams {
        agentId: string;
        queryEmbedding: readonly number[];
        threshold: number;
        limit: number;
        relevanceWeight: number;
        recencyWeight: number;
        queryText?: string;
        contentTypes?: readonly ContentType[];
    }
    Index
    agentId queryEmbedding threshold limit relevanceWeight recencyWeight queryText? contentTypes?
    agentId: string
    queryEmbedding: readonly number[]
    threshold: number
    limit: number
    relevanceWeight: number

    Blend weights for relevanceWeight·sim + recencyWeight·recency.

    recencyWeight: number
    queryText?: string

    Raw query text for hybrid retrieval. When provided, the threshold-admitted candidates are re-ranked by fusing embedding cosine with BM25 (lexical) and recency via RRF. Omit for the classic single-channel dense+recency blend.

    contentTypes?: readonly ContentType[]

    Restrict results to these content types. Omit for no filter. Matches the Python protocol's content_types, so one contract describes both libraries.

    Settings
    Member Visibility

    On This Page
    Properties