Milvus.Client.V2 2.6.0

Milvus C# SDK (V2) — Milvus.Client.V2

The second-generation .NET SDK for Milvus, the open-source vector database. It implements the request/response DTO API (mirroring the Java/C++/Rust V2 SDKs) over gRPC and targets net8.0, netstandard2.0 and net462.

Install

dotnet add package Milvus.Client.V2

Quickstart

using Milvus.Client.V2;
using Milvus.Client.V2.Requests.Collection;
using Milvus.Client.V2.Requests.Dml;
using Milvus.Client.V2.Requests.Dql;
using Milvus.Client.V2.Requests.Index;
using Milvus.Client.V2.Responses.Dql;
using Milvus.Client.V2.Types;

// Connect (see below for a token when authentication is enabled).
using var client = new MilvusClientV2(new ConnectConfig { Uri = "localhost:19530" });

// Build the collection schema with its fields.
var schema = new CollectionSchema { Name = "book" };
schema.Fields.Add(new FieldSchema("id", DataType.Int64, isPrimaryKey: true, autoId: true));
schema.Fields.Add(FieldSchema.CreateVarchar("title", maxLength: 256));
schema.Fields.Add(FieldSchema.CreateFloatVector("embedding", dimension: 128));

// Create the collection.
await client.CreateCollectionAsync(new CreateCollectionReq
{
    CollectionName = "book",
    Schema = schema
});

// Insert rows (row-based or column-based; here row-based).
await client.InsertAsync(new InsertReq
{
    CollectionName = "book",
    RowsData =
    [
        new Dictionary<string, object?>
        {
            ["title"] = "The Time Machine",
            ["embedding"] = Enumerable.Repeat(0.1f, 128).ToArray()
        }
    ]
});

// Build an index and load before searching.
await client.CreateIndexAsync(new CreateIndexReq
{
    CollectionName = "book",
    Indexes = [new IndexParam("embedding", indexType: IndexType.AutoIndex, metricType: SimilarityMetricType.Cosine)]
});
await client.LoadCollectionAsync(new LoadCollectionReq { CollectionName = "book" });

// Search.
SearchResp results = await client.SearchAsync(new SearchReq
{
    CollectionName = "book",
    VectorFieldName = "embedding",
    Vectors = [Enumerable.Repeat(0.1f, 128).ToArray()],
    MetricType = SimilarityMetricType.Cosine,
    Limit = 10
});

See docs/design/v2_design.md for the full design and the request/response DTO conventions, and the examples/v2/ project for runnable, per-feature sample programs (each one is a small console app that connects to a server, exercises one feature area, and cleans up after itself).

Connection

// Default port 19530, plaintext.
using var client = new MilvusClientV2(new ConnectConfig { Uri = "localhost:19530" });

// With authentication (username/password) and TLS (https:// scheme).
using var client = new MilvusClientV2(new ConnectConfig
{
    Uri = "https://host:19530",
    Username = "root",
    Password = "Milvus"
});

// With an API key instead.
using var client = new MilvusClientV2(new ConnectConfig { Uri = "https://host:19530", ApiKey = "..." });

// Connect to a specific database.
using var client = new MilvusClientV2(new ConnectConfig { Uri = "localhost:19530", Database = "my_db" });

The examples honor the MILVUS_URI and MILVUS_TOKEN environment variables (default localhost:19530 / none).

Documentation layout