Merge pull request #139 from drasticactions/embeddings-semantickernel

Add TextEmbedding for Semantic Kernel

LLama.Examples/LLama.Examples.csproj

@@ -27,6 +27,7 @@
   </PropertyGroup>
 
   <ItemGroup>
+    <PackageReference Include="Microsoft.Extensions.Logging.Console" Version="7.0.0" />
     <PackageReference Include="Microsoft.SemanticKernel" Version="0.21.230828.2-preview" />
   </ItemGroup>
 

LLama.Examples/NewVersion/SemanticKernelMemory.cs

@@ -0,0 +1,172 @@
+using Microsoft.SemanticKernel.Memory;
+using Microsoft.SemanticKernel;
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+using LLama.Common;
+using LLamaSharp.SemanticKernel.TextEmbedding;
+using Microsoft.SemanticKernel.AI.Embeddings;
+
+namespace LLama.Examples.NewVersion
+{
+    public class SemanticKernelMemory
+    {
+        private const string MemoryCollectionName = "SKGitHub";
+
+        public static async Task Run()
+        {
+            var loggerFactory = ConsoleLogger.LoggerFactory;
+            Console.WriteLine("Example from: https://github.com/microsoft/semantic-kernel/blob/main/dotnet/samples/KernelSyntaxExamples/Example14_SemanticMemory.cs");
+            Console.Write("Please input your model path: ");
+            var modelPath = Console.ReadLine();
+
+            var seed = 1337;
+            // Load weights into memory
+            var parameters = new ModelParams(modelPath)
+            {
+                Seed = seed,
+                EmbeddingMode = true
+            };
+
+            using var model = LLamaWeights.LoadFromFile(parameters);
+            var embedding = new LLamaEmbedder(model, parameters);
+
+            Console.WriteLine("====================================================");
+            Console.WriteLine("======== Semantic Memory (volatile, in RAM) ========");
+            Console.WriteLine("====================================================");
+
+            /* You can build your own semantic memory combining an Embedding Generator
+             * with a Memory storage that supports search by similarity (ie semantic search).
+             *
+             * In this example we use a volatile memory, a local simulation of a vector DB.
+             *
+             * You can replace VolatileMemoryStore with Qdrant (see QdrantMemoryStore connector)
+             * or implement your connectors for Pinecone, Vespa, Postgres + pgvector, SQLite VSS, etc.
+             */
+
+            var kernelWithCustomDb = Kernel.Builder
+                .WithLoggerFactory(ConsoleLogger.LoggerFactory)
+                .WithAIService<ITextEmbeddingGeneration>("local-llama-embed", new LLamaSharpEmbeddingGeneration(embedding), true)
+                .WithMemoryStorage(new VolatileMemoryStore())
+                .Build();
+
+            await RunExampleAsync(kernelWithCustomDb);
+        }
+
+        private static async Task RunExampleAsync(IKernel kernel)
+        {
+            await StoreMemoryAsync(kernel);
+
+            await SearchMemoryAsync(kernel, "How do I get started?");
+
+            /*
+            Output:
+
+            Query: How do I get started?
+
+            Result 1:
+              URL:     : https://github.com/microsoft/semantic-kernel/blob/main/README.md
+              Title    : README: Installation, getting started, and how to contribute
+
+            Result 2:
+              URL:     : https://github.com/microsoft/semantic-kernel/blob/main/samples/dotnet-jupyter-notebooks/00-getting-started.ipynb
+              Title    : Jupyter notebook describing how to get started with the Semantic Kernel
+
+            */
+
+            await SearchMemoryAsync(kernel, "Can I build a chat with SK?");
+
+            /*
+            Output:
+
+            Query: Can I build a chat with SK?
+
+            Result 1:
+              URL:     : https://github.com/microsoft/semantic-kernel/tree/main/samples/skills/ChatSkill/ChatGPT
+              Title    : Sample demonstrating how to create a chat skill interfacing with ChatGPT
+
+            Result 2:
+              URL:     : https://github.com/microsoft/semantic-kernel/blob/main/samples/apps/chat-summary-webapp-react/README.md
+              Title    : README: README associated with a sample chat summary react-based webapp
+
+            */
+
+            await SearchMemoryAsync(kernel, "Jupyter notebook");
+
+            await SearchMemoryAsync(kernel, "README: README associated with a sample chat summary react-based webapp");
+
+            await SearchMemoryAsync(kernel, "Jupyter notebook describing how to pass prompts from a file to a semantic skill or function");
+        }
+
+        private static async Task SearchMemoryAsync(IKernel kernel, string query)
+        {
+            Console.WriteLine("\nQuery: " + query + "\n");
+
+            var memories = kernel.Memory.SearchAsync(MemoryCollectionName, query, limit: 10, minRelevanceScore: 0.5);
+
+            int i = 0;
+            await foreach (MemoryQueryResult memory in memories)
+            {
+                Console.WriteLine($"Result {++i}:");
+                Console.WriteLine("  URL:     : " + memory.Metadata.Id);
+                Console.WriteLine("  Title    : " + memory.Metadata.Description);
+                Console.WriteLine("  Relevance: " + memory.Relevance);
+                Console.WriteLine();
+            }
+
+            Console.WriteLine("----------------------");
+        }
+
+        private static async Task StoreMemoryAsync(IKernel kernel)
+        {
+            /* Store some data in the semantic memory.
+             *
+             * When using Azure Cognitive Search the data is automatically indexed on write.
+             *
+             * When using the combination of VolatileStore and Embedding generation, SK takes
+             * care of creating and storing the index
+             */
+
+            Console.WriteLine("\nAdding some GitHub file URLs and their descriptions to the semantic memory.");
+            var githubFiles = SampleData();
+            var i = 0;
+            foreach (var entry in githubFiles)
+            {
+                var result = await kernel.Memory.SaveReferenceAsync(
+                    collection: MemoryCollectionName,
+                    externalSourceName: "GitHub",
+                    externalId: entry.Key,
+                    description: entry.Value,
+                    text: entry.Value);
+
+                Console.WriteLine($"#{++i} saved.");
+                Console.WriteLine(result);
+            }
+
+            Console.WriteLine("\n----------------------");
+        }
+
+        private static Dictionary<string, string> SampleData()
+        {
+            return new Dictionary<string, string>
+            {
+                ["https://github.com/microsoft/semantic-kernel/blob/main/README.md"]
+                    = "README: Installation, getting started, and how to contribute",
+                ["https://github.com/microsoft/semantic-kernel/blob/main/dotnet/notebooks/02-running-prompts-from-file.ipynb"]
+                    = "Jupyter notebook describing how to pass prompts from a file to a semantic skill or function",
+                ["https://github.com/microsoft/semantic-kernel/blob/main/dotnet/notebooks//00-getting-started.ipynb"]
+                    = "Jupyter notebook describing how to get started with the Semantic Kernel",
+                ["https://github.com/microsoft/semantic-kernel/tree/main/samples/skills/ChatSkill/ChatGPT"]
+                    = "Sample demonstrating how to create a chat skill interfacing with ChatGPT",
+                ["https://github.com/microsoft/semantic-kernel/blob/main/dotnet/src/SemanticKernel/Memory/VolatileMemoryStore.cs"]
+                    = "C# class that defines a volatile embedding store",
+                ["https://github.com/microsoft/semantic-kernel/blob/main/samples/dotnet/KernelHttpServer/README.md"]
+                    = "README: How to set up a Semantic Kernel Service API using Azure Function Runtime v4",
+                ["https://github.com/microsoft/semantic-kernel/blob/main/samples/apps/chat-summary-webapp-react/README.md"]
+                    = "README: README associated with a sample chat summary react-based webapp",
+            };
+        }
+    }
+}

LLama.Examples/NewVersion/TestRunner.cs

@@ -20,6 +20,7 @@ public static async Task Run()
             Console.WriteLine("10: Constrain response to json format using grammar.");
             Console.WriteLine("11: Semantic Kernel Prompt.");
             Console.WriteLine("12: Semantic Kernel Chat.");
+            Console.WriteLine("13: Semantic Kernel Memory.");
 
             while (true)
             {
@@ -78,6 +79,10 @@ public static async Task Run()
                 {
                     await SemanticKernelChat.Run();
                 }
+                else if (choice == 13)
+                {
+                    await SemanticKernelMemory.Run();
+                }
                 else
                 {
                     Console.WriteLine("Cannot parse your choice. Please select again.");

LLama.Examples/RepoUtils.cs

@@ -0,0 +1,40 @@
+using Microsoft.Extensions.Logging;
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace LLama.Examples
+{
+    /// <summary>
+    /// Basic logger printing to console
+    /// </summary>
+    internal static class ConsoleLogger
+    {
+        internal static ILogger Logger => LoggerFactory.CreateLogger<object>();
+
+        internal static ILoggerFactory LoggerFactory => s_loggerFactory.Value;
+
+        private static readonly Lazy<ILoggerFactory> s_loggerFactory = new(LogBuilder);
+
+        private static ILoggerFactory LogBuilder()
+        {
+            return Microsoft.Extensions.Logging.LoggerFactory.Create(builder =>
+            {
+                builder.SetMinimumLevel(LogLevel.Warning);
+
+                builder.AddFilter("Microsoft", LogLevel.Trace);
+                builder.AddFilter("Microsoft", LogLevel.Debug);
+                builder.AddFilter("Microsoft", LogLevel.Information);
+                builder.AddFilter("Microsoft", LogLevel.Warning);
+                builder.AddFilter("Microsoft", LogLevel.Error);
+
+                builder.AddFilter("Microsoft", LogLevel.Warning);
+                builder.AddFilter("System", LogLevel.Warning);
+
+                builder.AddConsole();
+            });
+        }
+    }
+}

LLama.SemanticKernel/README.md

@@ -6,6 +6,7 @@ For reference on how to implement it, view the following examples:
 
 - [SemanticKernelChat](../LLama.Examples/NewVersion/SemanticKernelChat.cs)
 - [SemanticKernelPrompt](../LLama.Examples/NewVersion/SemanticKernelPrompt.cs)
+- [SemanticKernelMemory](../LLama.Examples/NewVersion/SemanticKernelMemory.cs)
 
 ## ITextCompletion
 ```csharp
@@ -24,3 +25,14 @@ using var context = model.CreateContext(parameters);
 var ex = new InteractiveExecutor(context);
 var chatGPT = new LLamaSharpChatCompletion(ex);
 ```
+
+## ITextEmbeddingGeneration
+```csharp
+using var model = LLamaWeights.LoadFromFile(parameters);
+var embedding = new LLamaEmbedder(model, parameters);
+var kernelWithCustomDb = Kernel.Builder
+    .WithLoggerFactory(ConsoleLogger.LoggerFactory)
+    .WithAIService<ITextEmbeddingGeneration>("local-llama-embed", new LLamaSharpEmbeddingGeneration(embedding), true)
+    .WithMemoryStorage(new VolatileMemoryStore())
+    .Build();
+```

LLama.SemanticKernel/TextEmbedding/LLamaSharpEmbeddingGeneration.cs

@@ -0,0 +1,20 @@
+using LLama;
+using Microsoft.SemanticKernel.AI.Embeddings;
+
+namespace LLamaSharp.SemanticKernel.TextEmbedding;
+
+public sealed class LLamaSharpEmbeddingGeneration : ITextEmbeddingGeneration
+{
+    private LLamaEmbedder _embedder;
+
+    public LLamaSharpEmbeddingGeneration(LLamaEmbedder embedder)
+    {
+        _embedder = embedder;
+    }
+
+    /// <inheritdoc/>
+    public async Task<IList<ReadOnlyMemory<float>>> GenerateEmbeddingsAsync(IList<string> data, CancellationToken cancellationToken = default)
+    {
+        return data.Select(text => new ReadOnlyMemory<float>(_embedder.GetEmbeddings(text))).ToList();
+    }
+}