CSHARP-5717: Typed builders for vector indexes

evergreen/evergreen.yml

@@ -2099,7 +2099,7 @@ task_groups:
             - "AWS_SESSION_TOKEN"
           env:
             CLUSTER_PREFIX: dbx-csharp-search-index
-            MONGODB_VERSION: "7.0"
+            MONGODB_VERSION: "8.0"
           args:
             - ${DRIVERS_TOOLS}/.evergreen/atlas/setup-atlas-cluster.sh
       - command: expansions.update

src/MongoDB.Driver/CreateSearchIndexModel.cs

@@ -18,9 +18,10 @@
 namespace MongoDB.Driver
 {
     /// <summary>
-    /// Model for creating a search index.
+    /// Defines a vector search index model using a <see cref="BsonDocument"/> definition. Consider using
+    /// <see cref="CreateVectorIndexModel{TDocument}"/> to build Atlas indexes without specifying the BSON directly.
     /// </summary>
-    public sealed class CreateSearchIndexModel
+    public class CreateSearchIndexModel
     {
         /// <summary>Gets the index name.</summary>
         /// <value>The index name.</value>
@@ -32,26 +33,51 @@ public sealed class CreateSearchIndexModel
 
         /// <summary>Gets the index definition.</summary>
         /// <value>The definition.</value>
-        public BsonDocument Definition { get; }
+        public virtual BsonDocument Definition { get; }
 
         /// <summary>
-        /// Initializes a new instance of the <see cref="CreateSearchIndexModel"/> class.
+        /// Initializes a new instance of the <see cref="CreateSearchIndexModel"/> class, passing the index
+        /// model as a <see cref="BsonDocument"/>.
         /// </summary>
-        /// <param name="name">The name.</param>
-        /// <param name="definition">The definition.</param>
-        public CreateSearchIndexModel(string name, BsonDocument definition) : this(name, null, definition) { }
+        /// <remarks>
+        /// Consider using <see cref="CreateVectorIndexModel{TDocument}"/> to build Atlas indexes without specifying the
+        /// BSON directly.
+        /// </remarks>
+        /// <param name="name">The index name.</param>
+        /// <param name="definition">The index definition.</param>
+        public CreateSearchIndexModel(string name, BsonDocument definition)
+            : this(name, null, definition)
+        {
+        }
 
         /// <summary>
-        /// Initializes a new instance of the <see cref="CreateSearchIndexModel"/> class.
+        /// Initializes a new instance of the <see cref="CreateSearchIndexModel"/> class, passing the index
+        /// model as a <see cref="BsonDocument"/>.
         /// </summary>
-        /// <param name="name">The name.</param>
-        /// <param name="type">The type.</param>
-        /// <param name="definition">The definition.</param>
+        /// <remarks>
+        /// Consider using <see cref="CreateVectorIndexModel{TDocument}"/> to build indexes without specifying the
+        /// BSON directly.
+        /// </remarks>
+        /// <param name="name">The index name.</param>
+        /// <param name="type">The index type.</param>
+        /// <param name="definition">The index definition.</param>
         public CreateSearchIndexModel(string name, SearchIndexType? type, BsonDocument definition)
         {
             Name = name;
             Type = type;
             Definition = definition;
+
+        }
+
+        /// <summary>
+        /// Initializes a new instance of the <see cref="CreateSearchIndexModel"/> class.
+        /// </summary>
+        /// <param name="name">The index name.</param>
+        /// <param name="type">The index type.</param>
+        protected CreateSearchIndexModel(string name, SearchIndexType? type)
+        {
+            Name = name;
+            Type = type;
         }
     }
 }

src/MongoDB.Driver/CreateVectorIndexModel.cs

@@ -0,0 +1,173 @@
+/* Copyright 2010-present MongoDB Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Linq.Expressions;
+using MongoDB.Bson;
+
+namespace MongoDB.Driver;
+
+/// <summary>
+/// Defines a vector index model using strongly-typed C# APIs.
+/// </summary>
+public sealed class CreateVectorIndexModel<TDocument> : CreateSearchIndexModel
+{
+    /// <summary>
+    /// The field containing the vectors to index.
+    /// </summary>
+    public FieldDefinition<TDocument> Field { get; }
+
+    /// <summary>
+    /// The <see cref="VectorSimilarity"/> to use to search for top K-nearest neighbors.
+    /// </summary>
+    public VectorSimilarity Similarity { get; }
+
+    /// <summary>
+    /// Number of vector dimensions that vector search enforces at index-time and query-time.
+    /// </summary>
+    public int Dimensions { get; }
+
+    /// <summary>
+    /// Fields that may be used as filters in the vector query.
+    /// </summary>
+    public IReadOnlyList<FieldDefinition<TDocument>> FilterFields { get; }
+
+    /// <summary>
+    /// Type of automatic vector quantization for your vectors.
+    /// </summary>
+    public VectorQuantization? Quantization { get; init; }
+
+    /// <summary>
+    /// Maximum number of edges (or connections) that a node can have in the Hierarchical Navigable Small Worlds graph.
+    /// </summary>
+    public int? HnswMaxEdges { get; init; }
+
+    /// <summary>
+    /// Analogous to numCandidates at query-time, this parameter controls the maximum number of nodes to evaluate to find the closest neighbors to connect to a new node.
+    /// </summary>
+    public int? HnswNumEdgeCandidates { get; init; }
+
+    /// <summary>
+    /// This method should not be called on this subtype. Instead, call <see cref="Render"/> to create a BSON
+    /// document for the index model.
+    /// </summary>
+    public override BsonDocument Definition
+        => throw new NotSupportedException(
+            "This method should not be called on this subtype. Instead, call 'Render' to create a BSON document for the index model.");
+
+    /// <summary>
+    /// Initializes a new instance of the <see cref="CreateVectorIndexModel{TDocument}"/> class, passing the
+    /// required options for <see cref="VectorSimilarity"/> and the number of vector dimensions to the constructor.
+    /// </summary>
+    /// <param name="name">The index name.</param>
+    /// <param name="field">The field containing the vectors to index.</param>
+    /// <param name="similarity">The <see cref="VectorSimilarity"/> to use to search for top K-nearest neighbors.</param>
+    /// <param name="dimensions">Number of vector dimensions that vector search enforces at index-time and query-time.</param>
+    /// <param name="filterFields">Fields that may be used as filters in the vector query.</param>
+    public CreateVectorIndexModel(
+        FieldDefinition<TDocument> field,
+        string name,
+        VectorSimilarity similarity,
+        int dimensions,
+        params FieldDefinition<TDocument>[] filterFields)
+        : base(name, SearchIndexType.VectorSearch)
+    {
+        Field = field;
+        Similarity = similarity;
+        Dimensions = dimensions;
+        FilterFields = filterFields?.ToList() ?? [];
+    }
+
+    /// <summary>
+    /// Initializes a new instance of the <see cref="CreateVectorIndexModel{TDocument}"/> class, passing the
+    /// required options for <see cref="VectorSimilarity"/> and the number of vector dimensions to the constructor.
+    /// </summary>
+    /// <param name="name">The index name.</param>
+    /// <param name="field">An expression pointing to the field containing the vectors to index.</param>
+    /// <param name="similarity">The <see cref="VectorSimilarity"/> to use to search for top K-nearest neighbors.</param>
+    /// <param name="dimensions">Number of vector dimensions that vector search enforces at index-time and query-time.</param>
+    /// <param name="filterFields">Expressions pointing to fields that may be used as filters in the vector query.</param>
+    public CreateVectorIndexModel(
+        Expression<Func<TDocument, object>> field,
+        string name,
+        VectorSimilarity similarity,
+        int dimensions,
+        params Expression<Func<TDocument, object>>[] filterFields)
+        : this(
+            new ExpressionFieldDefinition<TDocument>(field),
+            name,
+            similarity,
+            dimensions,
+            filterFields?
+                .Select(f => (FieldDefinition<TDocument>)new ExpressionFieldDefinition<TDocument>(f))
+                .ToArray())
+    {
+    }
+
+    /// <summary>
+    /// Renders the index model to a <see cref="BsonDocument"/>.
+    /// </summary>
+    /// <param name="renderArgs">The render arguments.</param>
+    /// <returns>A <see cref="BsonDocument" />.</returns>
+    public BsonDocument Render(RenderArgs<TDocument> renderArgs)
+    {
+        var similarityValue = Similarity == VectorSimilarity.DotProduct
+            ? "dotProduct" // Because neither "DotProduct" or "dotproduct" are allowed.
+            : Similarity.ToString().ToLowerInvariant();
+
+        var vectorField = new BsonDocument
+        {
+            { "type", BsonString.Create("vector") },
+            { "path", Field.Render(renderArgs).FieldName },
+            { "numDimensions", BsonInt32.Create(Dimensions) },
+            { "similarity", BsonString.Create(similarityValue) },
+        };
+
+        if (Quantization.HasValue)
+        {
+            vectorField.Add("quantization", BsonString.Create(Quantization.ToString()?.ToLower()));
+        }
+
+        if (HnswMaxEdges != null || HnswNumEdgeCandidates != null)
+        {
+            var hnswDocument = new BsonDocument
+            {
+                { "maxEdges", BsonInt32.Create(HnswMaxEdges ?? 16) },
+                { "numEdgeCandidates", BsonInt32.Create(HnswNumEdgeCandidates ?? 100) }
+            };
+            vectorField.Add("hnswOptions", hnswDocument);
+        }
+
+        var fieldDocuments = new List<BsonDocument> { vectorField };
+
+        if (FilterFields != null)
+        {
+            foreach (var filterPath in FilterFields)
+            {
+                var fieldDocument = new BsonDocument
+                {
+                    { "type", BsonString.Create("filter") },
+                    { "path", BsonString.Create(filterPath.Render(renderArgs).FieldName) }
+                };
+
+                fieldDocuments.Add(fieldDocument);
+            }
+        }
+
+        return new BsonDocument { { "fields", BsonArray.Create(fieldDocuments) } };
+    }
+}

src/MongoDB.Driver/MongoCollectionImpl.cs

@@ -1741,10 +1741,22 @@ private PipelineDefinition<TDocument, BsonDocument> CreateListIndexesStage(strin
                 return new BsonDocumentStagePipelineDefinition<TDocument, BsonDocument>(new[] { stage });
             }
 
-            private CreateSearchIndexesOperation CreateCreateIndexesOperation(IEnumerable<CreateSearchIndexModel> models) =>
-                new(_collection._collectionNamespace,
-                    models.Select(m => new CreateSearchIndexRequest(m.Name, m.Type, m.Definition)),
+            private CreateSearchIndexesOperation CreateCreateIndexesOperation(
+                IEnumerable<CreateSearchIndexModel> models)
+            {
+                var renderArgs = _collection.GetRenderArgs();
+
+                return new CreateSearchIndexesOperation(
+                    _collection._collectionNamespace,
+                    models.Select(model
+                        => new CreateSearchIndexRequest(
+                            model.Name,
+                            model.Type,
+                            model is CreateVectorIndexModel<TDocument> createVectorIndexModel
+                                ? createVectorIndexModel.Render(renderArgs)
+                                : model.Definition)),
                     _collection._messageEncoderSettings);
+            }
 
             private string[] GetIndexNames(BsonDocument createSearchIndexesResponse) =>
                 createSearchIndexesResponse["indexesCreated"]

src/MongoDB.Driver/VectorQuantization.cs

@@ -0,0 +1,42 @@
+/* Copyright 2010-present MongoDB Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+namespace MongoDB.Driver;
+
+/// <summary>
+/// Type of automatic vector quantization for your vectors. Use this setting only if your embeddings are float
+/// or double vectors. See <see href="https://www.mongodb.com/docs/atlas/atlas-vector-search/vector-quantization/">
+/// Vector Quantization</see> for more information.
+/// </summary>
+public enum VectorQuantization
+{
+    /// <summary>
+    /// Indicates no automatic quantization for the vector embeddings. Use this setting if you have pre-quantized
+    /// vectors for ingestion. If omitted, this is the default value.
+    /// </summary>
+    None,
+
+    /// <summary>
+    /// Indicates scalar quantization, which transforms values to 1 byte integers.
+    /// </summary>
+    Scalar,
+
+    /// <summary>
+    /// Indicates binary quantization, which transforms values to a single bit.
+    /// To use this value, numDimensions must be a multiple of 8.
+    /// If precision is critical, select <see cref="None"/> or <see cref="Scalar"/> instead of <see cref="Binary"/>.
+    /// </summary>
+    Binary,
+}

src/MongoDB.Driver/VectorSimilarity.cs

@@ -0,0 +1,39 @@
+/* Copyright 2010-present MongoDB Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+namespace MongoDB.Driver;
+
+/// <summary>
+/// Vector similarity function to use to search for top K-nearest neighbors.
+/// See <see href="https://www.mongodb.com/docs/atlas/atlas-vector-search/vector-search-type/">How to Index Fields for
+/// Vector Search</see> for more information.
+/// </summary>
+public enum VectorSimilarity
+{
+    /// <summary>
+    /// Measures the distance between ends of vectors.
+    /// </summary>
+    Euclidean,
+
+    /// <summary>
+    /// Measures similarity based on the angle between vectors.
+    /// </summary>
+    Cosine,
+
+    /// <summary>
+    /// Measures similarity like cosine, but takes into account the magnitude of the vector.
+    /// </summary>
+    DotProduct,
+}