feat: add Azure Embeddings Interface implementation (#116)

LangChain.sln

@@ -465,6 +465,7 @@ Global
 		{B953ABEC-50DD-4A63-A12A-E82F124C7D5B} = {FDEE2E22-C239-4921-83B2-9797F765FD6A}
 		{32FC123E-F269-4352-848C-0161B53093CC} = {E55391DE-F8F3-4CC2-A0E3-2406C76E9C68}
 		{DEAFA0CB-462D-4D74-B16F-68FD83FE3858} = {FDEE2E22-C239-4921-83B2-9797F765FD6A}
+		{18F5AAB1-1750-41BD-B623-6339CA5754D9} = {E55391DE-F8F3-4CC2-A0E3-2406C76E9C68}
 		{72B1E2CC-1A34-470E-A579-034CB0972BB7} = {FDEE2E22-C239-4921-83B2-9797F765FD6A}
 		{4913844F-74EC-4E74-AE8A-EA825569E6BA} = {E55391DE-F8F3-4CC2-A0E3-2406C76E9C68}
 		{BF4C7B87-0997-4208-84EF-D368DF7B9861} = {E55391DE-F8F3-4CC2-A0E3-2406C76E9C68}

src/libs/Providers/LangChain.Providers.Azure/AzureOpenAIModel.Embeddings.cs

@@ -0,0 +1,125 @@
+using Azure;
+using Azure.AI.OpenAI;
+using OpenAI;
+using OpenAI.Constants;
+using OpenAI.Embeddings;
+using System;
+using System.Collections.Generic;
+using System.Diagnostics;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+using static System.Net.Mime.MediaTypeNames;
+using OpenAIClient = Azure.AI.OpenAI.OpenAIClient;
+
+namespace LangChain.Providers.Azure;
+
+public partial class AzureOpenAIModel : IEmbeddingModel
+{
+    #region Properties
+
+    /// <inheritdoc cref="OpenAiConfiguration.EmbeddingModelId"/>
+    public string EmbeddingModelId { get; init; } = EmbeddingModel.Ada002;
+
+    /// <summary>
+    /// API has limit of 2048 elements in array per request
+    /// so we need to split texts into batches
+    /// https://platform.openai.com/docs/api-reference/embeddings
+    /// </summary>
+    public int EmbeddingBatchSize { get; init; } = 2048;
+
+    /// <inheritdoc/>
+    public int MaximumInputLength => ContextLengths.Get(EmbeddingModelId);
+
+    #endregion
+
+    #region Methods
+
+    /// <inheritdoc/>
+    public async Task<float[]> EmbedQueryAsync(
+        string text,
+        CancellationToken cancellationToken = default)
+    {
+        var watch = Stopwatch.StartNew();
+
+        var embeddingOptions = new EmbeddingsOptions(Id, new[] { text });
+
+        var response = await Client.GetEmbeddingsAsync(embeddingOptions, cancellationToken).ConfigureAwait(false);
+
+        var usage = GetUsage(response) with
+        {
+            Time = watch.Elapsed,
+        };
+        lock (_usageLock)
+        {
+            TotalUsage += usage;
+        }
+
+        return response.Value.Data[0].Embedding.ToArray();
+    }
+
+    /// <inheritdoc/>
+    public async Task<float[][]> EmbedDocumentsAsync(
+        string[] texts,
+        CancellationToken cancellationToken = default)
+    {
+        texts = texts ?? throw new ArgumentNullException(nameof(texts));
+
+        var watch = Stopwatch.StartNew();
+
+        var index = 0;
+        var batches = new List<string[]>();
+        while (index < texts.Length)
+        {
+            batches.Add(texts.Skip(index).Take(EmbeddingBatchSize).ToArray());
+            index += EmbeddingBatchSize;
+        }
+
+        var results = await Task.WhenAll(batches.Select(async batch =>
+        {
+            var watch = Stopwatch.StartNew();
+            var embeddingOptions = new EmbeddingsOptions(Id, batch);
+
+            var response = await Client.GetEmbeddingsAsync(embeddingOptions, cancellationToken).ConfigureAwait(false);
+
+            var usage = GetUsage(response) with
+            {
+                Time = watch.Elapsed,
+            };
+            lock (_usageLock)
+            {
+                TotalUsage += usage;
+            }
+
+            return response.Value.Data
+                .Select(x => x.Embedding.ToArray())
+                .ToArray();
+        })).ConfigureAwait(false);
+
+        var rr = results
+            .SelectMany(x => x.ToArray())
+            .ToArray();
+        return rr;
+    }
+
+
+
+    private Usage GetUsage(Response<Embeddings>? response)
+    {
+        if (response?.Value?.Usage == null!)
+        {
+            return Usage.Empty;
+        }
+        var tokens = response.Value?.Usage.PromptTokens ?? 0;
+        var priceInUsd = EmbeddingPrices.TryGet(
+            model: new EmbeddingModel(EmbeddingModelId),
+            tokens: tokens) ?? 0.0D;
+
+        return Usage.Empty with
+        {
+            InputTokens = tokens,
+            PriceInUsd = priceInUsd,
+        };
+    }
+    #endregion
+}

src/libs/Providers/LangChain.Providers.Azure/AzureOpenAIModel.cs

@@ -7,8 +7,9 @@ namespace LangChain.Providers.Azure;
 /// <summary>
 /// Wrapper around Azure OpenAI large language models
 /// </summary>
-public class AzureOpenAIModel : IChatModel
+public partial class AzureOpenAIModel : IChatModel
 {
+    private readonly object _usageLock = new();
     /// <summary>
     /// Azure OpenAI API Key
     /// </summary>
@@ -29,7 +30,7 @@ public class AzureOpenAIModel : IChatModel
     /// Azure OpenAI Resource URI
     /// </summary>
     public string Endpoint { get; set; }
-
+    public OpenAIClient Client { get; set; }
     private AzureOpenAIConfiguration Configurations { get; }
 
     #region Constructors
@@ -46,6 +47,7 @@ public AzureOpenAIModel(string apiKey, string endpoint, string id)
         Id = id ?? throw new ArgumentNullException(nameof(id));
         ApiKey = apiKey ?? throw new ArgumentNullException(nameof(apiKey));
         Endpoint = endpoint ?? throw new ArgumentNullException(nameof(endpoint));
+        Client = new OpenAIClient(new Uri(Endpoint), new AzureKeyCredential(ApiKey));
     }
 
     /// <summary>
@@ -60,6 +62,7 @@ public AzureOpenAIModel(AzureOpenAIConfiguration configuration)
         ApiKey = configuration.ApiKey ?? throw new ArgumentException("ApiKey is not defined", nameof(configuration));
         Id = configuration.Id ?? throw new ArgumentException("Deployment model Id is not defined", nameof(configuration));
         Endpoint = configuration.Endpoint ?? throw new ArgumentException("Endpoint is not defined", nameof(configuration));
+        Client = new OpenAIClient(new Uri(Endpoint), new AzureKeyCredential(ApiKey));
     }
     #endregion
 
@@ -94,9 +97,8 @@ private async Task<Response<ChatCompletions>> CreateChatCompleteAsync(IReadOnlyC
             ChoiceCount = Configurations.ChoiceCount,
             Temperature = Configurations.Temperature,
         };
-
-        var client = new OpenAIClient(new Uri(Endpoint), new AzureKeyCredential(ApiKey));
-        return await client.GetChatCompletionsAsync(chatCompletionOptions, cancellationToken).ConfigureAwait(false);
+
+        return await Client.GetChatCompletionsAsync(chatCompletionOptions, cancellationToken).ConfigureAwait(false);
     }
 
     private static ChatRequestMessage ToRequestMessage(Message message)