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

redis-vl-dotnet

VectorRangeQuery

VectorRangeQuery performs radius-style vector search instead of KNN ranking.

When to use it

Use VectorRangeQuery when the main constraint is a maximum distance threshold:

var nearbyResults = await index.SearchAsync(
    VectorRangeQuery.FromFloat32(
        "embedding",
        queryVector,
        distanceThreshold: 0.25,
        returnFields: ["title", "vector_distance"],
        runtimeOptions: new VectorRangeRuntimeOptions(epsilon: 0.01)));

This behavior is covered directly in tests/RedisVL.Tests/Indexes/SearchQueryCommandBuilderTests.cs and in the semantic cache/message history workflows that use vector range matching internally.

Behavior

VectorRangeQuery differs from VectorQuery in a few important ways:

  • it requires distanceThreshold > 0

  • it does not use topK

  • standard pagination applies with no extra topK-window validation

  • scoreAlias still defaults to vector_distance

Runtime range options

Use VectorRangeRuntimeOptions for range-specific tuning:

  • epsilon cannot be negative

  • epsilon must be finite when provided

epsilon is passed at runtime and is separate from the index schema definition.

Paging and mapping

returnFields are normalized the same way as other search queries, and typed mapping works through the same SearchAsync<TDocument> and SearchBatchesAsync<TDocument> APIs.

Related pages