SubSearchRequest

This class represents a single ANN sub-search within a HybridSearch operation. Add target vectors and set search parameters using the fluent With/Add methods, then pass a SubSearchRequestPtr to HybridSearchRequest::AddSubRequest(). SubSearchRequest inherits the full vector-assigning API from SearchRequestVectorAssigner.

SubSearchRequest req;
SubSearchRequestPtr ptr = std::make_shared<SubSearchRequest>(std::move(req));

using SubSearchRequestPtr = std::shared_ptr<SubSearchRequest>;

Request Syntax

SubSearchRequest()
    .WithAnnsField(field_name)
    .WithLimit(limit)
    .WithFilter(filter)
    .WithMetricType(metric_type)
    .WithTimezone(tz)
    .AddFloatVector(vector)       // or any Add*/With* vector method
    .WithFloatVectors(vectors);   // batch assignment

REQUEST METHODS:

• SubSearchRequest& WithAnnsField(const std::string& ann_field)

  Target vector field name for this sub-search.

• SubSearchRequest& WithLimit(int64_t limit)

  Top-k limit for this sub-search result before reranking. This value is stored in extra params.

• SubSearchRequest& WithFilter(std::string filter)

  Boolean filter expression applied to this sub-search only.

• SubSearchRequest& WithMetricType(milvus::MetricType metric_type)

  Metric type override for this sub-search.

• SubSearchRequest& WithTimezone(const std::string& timezone)

  Timezone string for Timestamptz field filtering.

Inherited vector methods (all return SubSearchRequest& for chaining):

• AddFloatVector(const FloatVecFieldData::ElementT& vector)

• AddBinaryVector(const std::string& vector)

• AddSparseVector(const SparseFloatVecFieldData::ElementT& vector)

• AddFloat16Vector(const Float16VecFieldData::ElementT& vector)

• AddBFloat16Vector(const BFloat16VecFieldData::ElementT& vector)

• AddInt8Vector(const Int8VecFieldData::ElementT& vector)

• AddEmbeddedText(const std::string& text)

• AddEmbeddingList(EmbeddingList&& emb_list) — for struct-field ANN

• WithFloatVectors(std::vector<FloatVecFieldData::ElementT>&& vectors) — batch

• WithSparseVectors(...), WithFloat16Vectors(...), etc. — batch variants

Example

#include <milvus/MilvusClientV2.h>
using namespace milvus;

auto client = MilvusClientV2::Create();
client->Connect(ConnectParam("http://localhost:19530").WithToken("root:Milvus"));

auto sub_dense = SubSearchRequest()
    .WithAnnsField("dense_vec")
    .WithLimit(20)
    .WithFilter("category == \"electronics\"")
    .AddFloatVector(std::vector<float>(128, 0.1f));

auto sub_sparse = SubSearchRequest()
    .WithAnnsField("sparse_vec")
    .WithLimit(20)
    .AddSparseVector({{0u, 0.3f}, {7u, 0.5f}});

auto reranker = std::make_shared<RRFRerank>(60);

SearchResponse response;
auto status = client->HybridSearch(
    HybridSearchRequest()
        .WithCollectionName("my_collection")
        .WithLimit(10)
        .AddSubRequest(std::make_shared<SubSearchRequest>(std::move(sub_dense)))
        .AddSubRequest(std::make_shared<SubSearchRequest>(std::move(sub_sparse)))
        .WithRerank(reranker),
    response);