Redis
redis-vl-dotnet Documentation redis-vl-dotnet

redis-vl-dotnet

AggregateHybridQuery

AggregateHybridQuery applies a hybrid text-plus-vector match and then runs an aggregation pipeline over the matched documents.

Basic shape

var aggregateHybridResults = await index.AggregateAsync<GenreHybridSummary>(
    AggregateHybridQuery.FromFloat32(
        Filter.Text("title").Prefix("He") | Filter.Text("title").Prefix("Ar"),
        "plot_embedding",
        queryVector,
        topK: 3,
        groupBy: new AggregationGroupBy(
            ["genre"],
            [
                AggregationReducer.Count("matchCount"),
                AggregationReducer.Average("vector_distance", "avgDistance")
            ]),
        sortBy: new AggregationSortBy(
            [
                new AggregationSortField("matchCount", descending: true),
                new AggregationSortField("avgDistance")
            ]),
        runtimeOptions: new VectorKnnRuntimeOptions(efRuntime: 90)));

The canonical runnable reference is /examples/VectorSearchExample.

Key behavior

AggregateHybridQuery combines the rules from hybrid search and aggregation:

  • textFilter must contain at least one text predicate

  • topK defines the hybrid retrieval window before aggregation runs

  • loadFields are normalized and deduplicated, preserving explicit @field syntax when provided

  • scoreAlias defaults to vector_distance

  • VectorKnnRuntimeOptions controls the runtime EF_RUNTIME parameter

Unlike HybridQuery, aggregate paging applies to the aggregation rows through the normal offset and limit values.

Typed row mapping and batches

Use AggregateAsync<TDocument> to map rows into projection types and AggregateBatchesAsync<TDocument> to stream multi-page result sets. Both behaviors are covered in tests/RedisVL.Tests/Indexes/SearchIndexAsyncTests.cs.

Related pages