メインコンテンツまでスキップ

hybridSearch()

Addedv2.4.xModifiedv2.6.x

This operation conducts a hybrid search across multiple vector fields with an optional scalar filtering expression and returns the merged, reranked results.

await milvusClient.hybridSearch(data: HybridSearchReq)

Request Syntax​

await milvusClient.hybridSearch({
    collection_name: string,
    data: HybridSearchSingleReq[],
    limit?: number,
    offset?: number,
    output_fields?: string[],
    filter?: string,
    rerank?: RerankerObj | FunctionObject | FunctionScore,
    partition_names?: string[],
    metric_type?: string,
    consistency_level?: ConsistencyLevelEnum,
    ignore_growing?: boolean,
    group_by_field?: string,
    group_size?: number,
    strict_group_size?: boolean,
    hints?: string,
    round_decimal?: number,
    transformers?: OutputTransformers,
    db_name?: string,
    timeout?: number,
})

PARAMETERS:

  • collection_name (string) -

    [REQUIRED]

    The name of the collection to search.

  • data (HybridSearchSingleReq[]) -

    [REQUIRED]

    A list of sub-search requests, one per vector field. Each element defines the query vector and target field for a single-vector sub-search. For the full field reference, see the HybridSearchSingleReq section below.

  • limit (number) -

    The total number of entities to return. The sum of this value and offset must be less than 16,384.

  • offset (number) -

    The number of records to skip in the search result. The sum of this value and limit must be less than 16,384.

  • output_fields (string[]) -

    A list of field names to include in each returned entity. Only the primary field is included by default.

  • filter (string) -

    A top-level scalar filtering condition applied after the hybrid search results are merged. Defaults to an empty string.

  • rerank (RerankerObj | FunctionObject | FunctionScore) -

    A reranking strategy for combining results from multiple sub-searches. See search() for the full rerank parameter schema.

  • partition_names (string[]) -

    The names of the partitions to search.

  • metric_type (string) -

    The metric type used to measure similarity between vectors.

  • consistency_level (ConsistencyLevelEnum) -

    The consistency level of the target collection. Options: Strong (0), Bounded (1), Session (2), Eventually (3). Defaults to Bounded.

  • ignore_growing (boolean) -

    Whether to skip growing segments during the search.

  • group_by_field (string) -

    Groups search results by the specified field to ensure diversity and avoid returning multiple results from the same group.

  • group_size (number) -

    The target number of entities to return within each group in a grouping search.

  • strict_group_size (boolean) -

    Whether to strictly enforce group_size. When true, the system attempts to fill each group with exactly group_size results.

  • hints (string) -

    A hints string to improve search performance.

  • round_decimal (number) -

    The number of decimal places to keep in the final scores.

  • transformers (OutputTransformers) -

    Custom transformers for special vector data types such as BFloat16Vector and Float16Vector.

  • db_name (string) -

    The name of the database containing the collection.

  • timeout (number) -

    The timeout duration for this operation in milliseconds.

RETURNS:

Promise<SearchResults>

This method returns a promise that resolves to a SearchResults object.

EXCEPTIONS:

  • MilvusError

    This exception will be raised when any error occurs during this operation.

HybridSearchSingleReq​

Each element in the data array is a HybridSearchSingleReq object that defines a single-vector sub-search request.

PARAMETERS:

  • data (SearchData) -

    [REQUIRED]

    The query vector for this sub-search. Can be a dense vector (number[]), a sparse vector (SparseVectorDic), or a text string for text-based search.

  • anns_field (string) -

    [REQUIRED]

    The name of the vector field to search within this sub-request.

  • filter (string) -

    A scalar filtering condition applied only to this sub-search.

  • exprValues (keyValueObj) -

    Template values for the filter expression in key-value pairs.

  • params (keyValueObj) -

    Index-specific search parameters in key-value pairs.

  • ignore_growing (boolean) -

    Whether to skip growing segments during this sub-search.

  • group_by_field (string) -

    Groups results by the specified field to ensure diversity within this sub-search.

  • transformers (OutputTransformers) -

    Custom transformers for special vector types such as BFloat16Vector and Float16Vector.

Example​

import { MilvusClient } from '@zilliz/milvus2-sdk-node';

const milvusClient = new MilvusClient({
    address: 'YOUR_CLUSTER_ENDPOINT',
    token: 'YOUR_CLUSTER_TOKEN',
});

const results = await milvusClient.hybridSearch({
    collection_name: 'my_collection',
    data: [
        {
            anns_field: 'dense_vector',
            data: [0.1, 0.2, 0.3, 0.4, 0.5],
        },
        {
            anns_field: 'sparse_vector',
            data: { 1: 0.5, 42: 0.8, 100: 0.3 },
        },
    ],
    limit: 10,
    rerank: { strategy: 'rrf', params: { k: 60 } },
    output_fields: ['id', 'text'],
});

console.log(results.results);