v2.1.3

src/Build5Nines.SharpVector.OpenAI/Embeddings/OpenAIEmbeddingsGenerator.cs

@@ -1,9 +1,11 @@
 using Build5Nines.SharpVector.Embeddings;
 using OpenAI.Embeddings;
+using System.Collections.Generic;
+using System.Linq;
 
 namespace Build5Nines.SharpVector.OpenAI.Embeddings;
 
-public class OpenAIEmbeddingsGenerator : IEmbeddingsGenerator
+public class OpenAIEmbeddingsGenerator :  IEmbeddingsGenerator //IBatchEmbeddingsGenerator
 {
     protected EmbeddingClient EmbeddingClient { get; private set; }
 
@@ -18,4 +20,29 @@ public async Task<float[]> GenerateEmbeddingsAsync(string text)
         var vector = embedding.ToFloats();
         return vector.ToArray();
     }
+
+    /// <summary>
+    /// Generates embeddings for a batch of input texts using the OpenAI embeddings client.
+    /// This leverages the API's multi-input batching for improved throughput and reduced overhead.
+    /// </summary>
+    /// <param name="texts">Collection of non-empty texts to embed.</param>
+    /// <returns>A list of float vectors aligned to the input order.</returns>
+    public async Task<IReadOnlyList<float[]>> GenerateEmbeddingsAsync(IEnumerable<string> texts)
+    {
+        if (texts is null) throw new ArgumentNullException(nameof(texts));
+
+        var inputs = texts.ToList();
+        if (inputs.Count == 0)
+        {
+            return Array.Empty<float[]>();
+        }
+
+        // Call the batch embeddings API once for all inputs.
+        var batchResult = await EmbeddingClient.GenerateEmbeddingsAsync(inputs);
+
+        // Map the embeddings to float arrays while preserving order.
+        var vectors = batchResult.Value.Select(e => e.ToFloats().ToArray()).ToList();
+
+        return vectors;
+    }
 }

src/Build5Nines.SharpVector/Build5Nines.SharpVector.csproj

@@ -9,7 +9,7 @@
     <PackageId>Build5Nines.SharpVector</PackageId>
     <PackageProjectUrl>https://sharpvector.build5nines.com</PackageProjectUrl>
     <RepositoryUrl>https://github.com/Build5Nines/SharpVector</RepositoryUrl>
-    <Version>2.1.2</Version>
+    <Version>2.1.3</Version>
     <Description>Lightweight In-memory Vector Database to embed in any .NET Applications</Description>
     <Copyright>Copyright (c) 2025 Build5Nines LLC</Copyright>
     <PackageReadmeFile>README.md</PackageReadmeFile>

src/Build5Nines.SharpVector/Embeddings/IBatchEmbeddingsGenerator.cs

@@ -0,0 +1,15 @@
+namespace Build5Nines.SharpVector.Embeddings;
+
+/// <summary>
+/// Optional capability for embeddings generators to support batch embedding of multiple texts.
+/// Implementations can leverage provider APIs that accept multi-input requests for better performance.
+/// </summary>
+public interface IBatchEmbeddingsGenerator : IEmbeddingsGenerator
+{
+    /// <summary>
+    /// Generates embeddings for multiple input texts in a single call when supported.
+    /// </summary>
+    /// <param name="texts">Collection of texts to embed. Order should be preserved in output.</param>
+    /// <returns>A read-only list of embeddings vectors corresponding to the input order.</returns>
+    Task<IReadOnlyList<float[]>> GenerateEmbeddingsAsync(IEnumerable<string> texts);
+}

src/Build5Nines.SharpVector/IVectorDatabase.cs

@@ -28,6 +28,13 @@ public interface IVectorDatabase<TId, TMetadata, TDocument>
     /// <returns></returns>
     Task<TId> AddTextAsync(TDocument text, TMetadata? metadata = default(TMetadata));
 
+    /// <summary>
+    /// Adds multiple texts with Metadata to the database and returns their IDs
+    /// </summary>
+    /// <param name="items">The texts and metadata to add in batch.</param>
+    /// <returns>The IDs of the added texts.</returns>
+    Task<IReadOnlyList<TId>> AddTextsAsync(IEnumerable<(TDocument text, TMetadata? metadata)> items);
+
     /// <summary>
     /// Get all the Ids for each text the database.
     /// </summary>

src/Build5Nines.SharpVector/MemoryVectorDatabaseBase.cs

@@ -107,6 +107,22 @@ public IEnumerable<TId> GetIds()
         return id;
     }
 
+    public async Task<IReadOnlyList<TId>> AddTextsAsync(IEnumerable<(TVocabularyKey text, TMetadata? metadata)> items)
+    {
+        if (items is null) throw new ArgumentNullException(nameof(items));
+
+        var ids = new List<TId>();
+
+        foreach(var item in items)
+        {
+            TId id = await AddTextAsync(item.text, item.metadata);
+            ids.Add(id);
+        }
+
+        return ids;
+    }
+
+
     /// <summary>
     /// Retrieves a text and metadata by its ID
     /// </summary>
@@ -469,6 +485,48 @@ public IEnumerable<TId> GetIds()
         return id;
     }
 
+    /// <summary>
+    /// Adds multiple texts with optional metadata to the database efficiently.
+    /// If the embeddings generator supports batching, this will generate vectors in a single multi-input call.
+    /// </summary>
+    /// <param name="items">Collection of (text, metadata) tuples to add.</param>
+    /// <returns>List of generated IDs in the same order as inputs.</returns>
+    public async Task<IReadOnlyList<TId>> AddTextsAsync(IEnumerable<(string text, TMetadata? metadata)> items)
+    {
+        if (items is null) throw new ArgumentNullException(nameof(items));
+
+        var list = items.ToList();
+        if (list.Count == 0) return Array.Empty<TId>();
+
+        // Try batch embeddings if supported
+        float[][] vectors;
+        if (EmbeddingsGenerator is IBatchEmbeddingsGenerator batchGen)
+        {
+            var batch = await batchGen.GenerateEmbeddingsAsync(list.Select(i => i.text));
+            vectors = batch.Select(v => v.ToArray()).ToArray();
+        }
+        else
+        {
+            // Fallback to per-item embedding
+            vectors = new float[list.Count][];
+            for (int i = 0; i < list.Count; i++)
+            {
+                vectors[i] = await EmbeddingsGenerator.GenerateEmbeddingsAsync(list[i].text);
+            }
+        }
+
+        // Store items and produce IDs
+        var ids = new List<TId>(list.Count);
+        for (int i = 0; i < list.Count; i++)
+        {
+            TId id = _idGenerator.NewId();
+            ids.Add(id);
+            await VectorStore.SetAsync(id, new VectorTextItem<TMetadata>(list[i].text, list[i].metadata, vectors[i]));
+        }
+
+        return ids;
+    }
+
     /// <summary>
     /// Retrieves a text and metadata by its ID
     /// </summary>

src/OpenAIConsoleTest/OpenAIConsoleTest.csproj

@@ -14,6 +14,7 @@
   </ItemGroup>
 
   <ItemGroup>
+    <!-- <ProjectReference Include="..\Build5Nines.SharpVector\Build5Nines.SharpVector.csproj" /> -->
     <ProjectReference Include="..\Build5Nines.SharpVector.OpenAI\Build5Nines.SharpVector.OpenAI.csproj" />
   </ItemGroup>
 

src/SharpVectorTest/BatchAddTests.cs

@@ -0,0 +1,74 @@
+namespace SharpVectorTest;
+
+using System.Linq;
+using System.Threading.Tasks;
+using Build5Nines.SharpVector;
+using Build5Nines.SharpVector.Embeddings;
+using Build5Nines.SharpVector.Id;
+using Build5Nines.SharpVector.VectorCompare;
+using Build5Nines.SharpVector.VectorStore;
+
+[TestClass]
+public class BatchAddTests
+{
+    [TestMethod]
+    public async Task AddTextsAsync_UsesBatchEmbeddings_WhenAvailable()
+    {
+        var db = new BatchMockMemoryVectorDatabase();
+
+        var inputs = new (string text, string? metadata)[]
+        {
+            ("one", "m1"),
+            ("two", "m2"),
+            ("three", "m3")
+        };
+
+        var ids = await db.AddTextsAsync(inputs);
+
+        Assert.AreEqual(3, ids.Count);
+
+        var results = db.Search("one");
+        Assert.AreEqual(3, results.Texts.Count());
+
+        // Ensure vectors were assigned from batch generator (length = 5 per mock)
+        foreach (var item in db)
+        {
+            Assert.AreEqual(5, item.Vector.Length);
+        }
+    }
+}
+
+public class BatchMockMemoryVectorDatabase
+     : MemoryVectorDatabaseBase<
+        int,
+        string,
+        MemoryDictionaryVectorStore<int, string>,
+        IntIdGenerator,
+        CosineSimilarityVectorComparer
+        >
+{
+    public BatchMockMemoryVectorDatabase()
+        : base(
+            new MockBatchEmbeddingsGenerator(),
+            new MemoryDictionaryVectorStore<int, string>()
+            )
+    { }
+}
+
+public class MockBatchEmbeddingsGenerator : IEmbeddingsGenerator, IBatchEmbeddingsGenerator
+{
+#pragma warning disable CS1998 // Async method lacks 'await' operators and will run synchronously
+    public async Task<float[]> GenerateEmbeddingsAsync(string text)
+#pragma warning restore CS1998 // Async method lacks 'await' operators and will run synchronously
+    {
+        return new float[] { 0.1f, 0.2f, 0.3f, 0.4f, 0.5f };
+    }
+
+#pragma warning disable CS1998 // Async method lacks 'await' operators and will run synchronously
+    public async Task<IReadOnlyList<float[]>> GenerateEmbeddingsAsync(IEnumerable<string> texts)
+#pragma warning restore CS1998 // Async method lacks 'await' operators and will run synchronously
+    {
+        // Return a different first value to ensure we can recognize batched path if needed
+        return texts.Select((t, idx) => new float[] { 0.9f, 0.2f, 0.3f, 0.4f, 0.5f }).ToList();
+    }
+}

src/SharpVectorTest/VectorDatabaseTests.cs

@@ -98,6 +98,30 @@ public void BasicMemoryVectorDatabase_05()
         Assert.AreEqual("metadata2", results.Texts.First().Metadata);
     }
 
+    [TestMethod]
+    public void BasicMemoryVectorDatabase_05_Batch()
+    {
+        var vdb = new BasicMemoryVectorDatabase();
+
+        // // Load Vector Database with some sample text
+        var inputs = new (string text, string? metadata)[]
+        {
+            ("The 👑 King", "metadata1"),
+            ("It's 🔥 Fire.", "metadata2"),
+            ("No emoji", "metadata3")
+        };
+        vdb.AddTextsAsync(inputs).Wait();
+
+        var results = vdb.Search("🔥", pageCount: 1);
+
+        Assert.AreEqual(1, results.Texts.Count());
+        Assert.AreEqual(0.5773503184318542, results.Texts.First().Similarity);
+        Assert.AreEqual("It's 🔥 Fire.", results.Texts.First().Text);
+        Assert.AreEqual(2, results.Texts.First().Id);
+        Assert.AreEqual("metadata2", results.Texts.First().Metadata);
+    }
+
+
     [TestMethod]
     public void BasicMemoryVectorDatabase_06()
     {