Support SemanticKernel 1.0.0-beta1

LLama.Examples/LLama.Examples.csproj

@@ -28,7 +28,7 @@
 
   <ItemGroup>
     <PackageReference Include="Microsoft.Extensions.Logging.Console" Version="7.0.0" />
-    <PackageReference Include="Microsoft.SemanticKernel" Version="0.21.230828.2-preview" />
+    <PackageReference Include="Microsoft.SemanticKernel" Version="1.0.0-beta1" />
   </ItemGroup>
 
   <ItemGroup>

LLama.Examples/NewVersion/SemanticKernelMemory.cs

@@ -1,13 +1,9 @@
-using Microsoft.SemanticKernel.Memory;
+using LLama.Common;
 using Microsoft.SemanticKernel;
-using System;
-using System.Collections.Generic;
-using System.Linq;
-using System.Text;
-using System.Threading.Tasks;
-using LLama.Common;
+using Microsoft.SemanticKernel.Memory;
 using LLamaSharp.SemanticKernel.TextEmbedding;
 using Microsoft.SemanticKernel.AI.Embeddings;
+using Microsoft.SemanticKernel.Plugins.Memory;
 
 namespace LLama.Examples.NewVersion
 {

LLama.Examples/NewVersion/SemanticKernelPrompt.cs

@@ -1,10 +1,7 @@
-using System.Reflection.Metadata;
-using System.Security.Cryptography;
-using System.Text;
-using LLama.Abstractions;
+using System.Security.Cryptography;
 using LLama.Common;
+using LLamaSharp.SemanticKernel.ChatCompletion;
 using Microsoft.SemanticKernel;
-using Microsoft.SemanticKernel.AI.ChatCompletion;
 using Microsoft.SemanticKernel.AI.TextCompletion;
 using LLamaSharp.SemanticKernel.TextCompletion;
 
@@ -35,7 +32,8 @@ namespace LLama.Examples.NewVersion
 
 One line TLDR with the fewest words.";
 
-            var summarize = kernel.CreateSemanticFunction(prompt, maxTokens: 100);
+            ChatRequestSettings settings = new() {MaxTokens = 100};
+            var summarize = kernel.CreateSemanticFunction(prompt, requestSettings: settings);
 
             string text1 = @"
 1st Law of Thermodynamics - Energy cannot be created or destroyed.
@@ -47,9 +45,10 @@ One line TLDR with the fewest words.";
 2. The acceleration of an object depends on the mass of the object and the amount of force applied.
 3. Whenever one object exerts a force on another object, the second object exerts an equal and opposite on the first.";
 
-            Console.WriteLine(await summarize.InvokeAsync(text1));
+            Console.WriteLine(await kernel.RunAsync(text1, summarize));
 
-            Console.WriteLine(await summarize.InvokeAsync(text2));
+            Console.WriteLine(await kernel.RunAsync(text2, summarize));
         }
     }
 }
+ 

LLama.SemanticKernel/ChatCompletion/ChatRequestSettings.cs

@@ -0,0 +1,54 @@
+using Microsoft.SemanticKernel.AI;
+
+namespace LLamaSharp.SemanticKernel.ChatCompletion;
+
+public class ChatRequestSettings : AIRequestSettings
+{
+    /// <summary>
+    /// Temperature controls the randomness of the completion.
+    /// The higher the temperature, the more random the completion.
+    /// </summary>
+    public double Temperature { get; set; } = 0;
+
+    /// <summary>
+    /// TopP controls the diversity of the completion.
+    /// The higher the TopP, the more diverse the completion.
+    /// </summary>
+    public double TopP { get; set; } = 0;
+
+    /// <summary>
+    /// Number between -2.0 and 2.0. Positive values penalize new tokens
+    /// based on whether they appear in the text so far, increasing the
+    /// model's likelihood to talk about new topics.
+    /// </summary>
+    public double PresencePenalty { get; set; } = 0;
+
+    /// <summary>
+    /// Number between -2.0 and 2.0. Positive values penalize new tokens
+    /// based on their existing frequency in the text so far, decreasing
+    /// the model's likelihood to repeat the same line verbatim.
+    /// </summary>
+    public double FrequencyPenalty { get; set; } = 0;
+
+    /// <summary>
+    /// Sequences where the completion will stop generating further tokens.
+    /// </summary>
+    public IList<string> StopSequences { get; set; } = Array.Empty<string>();
+
+    /// <summary>
+    /// How many completions to generate for each prompt. Default is 1.
+    /// Note: Because this parameter generates many completions, it can quickly consume your token quota.
+    /// Use carefully and ensure that you have reasonable settings for max_tokens and stop.
+    /// </summary>
+    public int ResultsPerPrompt { get; set; } = 1;
+
+    /// <summary>
+    /// The maximum number of tokens to generate in the completion.
+    /// </summary>
+    public int? MaxTokens { get; set; }
+
+    /// <summary>
+    /// Modify the likelihood of specified tokens appearing in the completion.
+    /// </summary>
+    public IDictionary<int, int> TokenSelectionBiases { get; set; } = new Dictionary<int, int>();
+}

LLama.SemanticKernel/ChatCompletion/LLamaSharpChatCompletion.cs

@@ -1,4 +1,5 @@
 using LLama;
+using Microsoft.SemanticKernel.AI;
 using Microsoft.SemanticKernel.AI.ChatCompletion;
 using System.Runtime.CompilerServices;
 
@@ -14,30 +15,29 @@ public sealed class LLamaSharpChatCompletion : IChatCompletion
     private ChatSession session;
     private ChatRequestSettings defaultRequestSettings;
 
-    public LLamaSharpChatCompletion(InteractiveExecutor model, ChatRequestSettings? defaultRequestSettings = default)
+    static ChatRequestSettings GetDefaultSettings()
     {
-        this.session = new ChatSession(model)
-            .WithHistoryTransform(new HistoryTransform())
-            .WithOutputTransform(new LLamaTransforms.KeywordTextOutputStreamTransform(new string[] { UserRole, AssistantRole }));
-        this.defaultRequestSettings = defaultRequestSettings ??= new ChatRequestSettings()
+        return new ChatRequestSettings
         {
             MaxTokens = 256,
             Temperature = 0,
             TopP = 0,
-            StopSequences = new List<string> { }
+            StopSequences = new List<string>()
         };
     }
 
+    public LLamaSharpChatCompletion(InteractiveExecutor model, ChatRequestSettings? defaultRequestSettings = default)
+    {
+        this.session = new ChatSession(model)
+            .WithHistoryTransform(new HistoryTransform())
+            .WithOutputTransform(new LLamaTransforms.KeywordTextOutputStreamTransform(new string[] { UserRole, AssistantRole }));
+        this.defaultRequestSettings = defaultRequestSettings ??= GetDefaultSettings();
+    }
+
     public LLamaSharpChatCompletion(ChatSession session, ChatRequestSettings? defaultRequestSettings = default)
     {
         this.session = session;
-        this.defaultRequestSettings = defaultRequestSettings ??= new ChatRequestSettings()
-        {
-            MaxTokens = 256,
-            Temperature = 0,
-            TopP = 0,
-            StopSequences = new List<string> { }
-        };
+        this.defaultRequestSettings = defaultRequestSettings ??= GetDefaultSettings();
     }
 
     /// <inheritdoc/>
@@ -54,21 +54,29 @@ public sealed class LLamaSharpChatCompletion : IChatCompletion
     }
 
     /// <inheritdoc/>
-    public async Task<IReadOnlyList<IChatResult>> GetChatCompletionsAsync(ChatHistory chat, ChatRequestSettings? requestSettings = null, CancellationToken cancellationToken = default)
+    public Task<IReadOnlyList<IChatResult>> GetChatCompletionsAsync(ChatHistory chat, AIRequestSettings? requestSettings = null, CancellationToken cancellationToken = default)
     {
-        requestSettings = requestSettings ?? this.defaultRequestSettings;
+        var settings = requestSettings != null 
+            ? (ChatRequestSettings)requestSettings
+            : defaultRequestSettings;
 
-        var result = this.session.ChatAsync(chat.ToLLamaSharpChatHistory(), requestSettings.ToLLamaSharpInferenceParams(), cancellationToken);
+        // This call is not awaited because LLamaSharpChatResult accepts an IAsyncEnumerable.
+        var result = this.session.ChatAsync(chat.ToLLamaSharpChatHistory(), settings.ToLLamaSharpInferenceParams(), cancellationToken);
 
-        return new List<IChatResult> { new LLamaSharpChatResult(result) }.AsReadOnly();
+        return Task.FromResult<IReadOnlyList<IChatResult>>(new List<IChatResult> { new LLamaSharpChatResult(result) }.AsReadOnly());
     }
 
     /// <inheritdoc/>
-    public async IAsyncEnumerable<IChatStreamingResult> GetStreamingChatCompletionsAsync(ChatHistory chat, ChatRequestSettings? requestSettings = null, [EnumeratorCancellation] CancellationToken cancellationToken = default)
+#pragma warning disable CS1998 // Async method lacks 'await' operators and will run synchronously.
+    public async IAsyncEnumerable<IChatStreamingResult> GetStreamingChatCompletionsAsync(ChatHistory chat, AIRequestSettings? requestSettings = null, [EnumeratorCancellation] CancellationToken cancellationToken = default)
+#pragma warning restore CS1998
     {
-        requestSettings = requestSettings ?? this.defaultRequestSettings;
+        var settings = requestSettings != null
+            ? (ChatRequestSettings)requestSettings
+            : defaultRequestSettings;
 
-        var result = this.session.ChatAsync(chat.ToLLamaSharpChatHistory(), requestSettings.ToLLamaSharpInferenceParams(), cancellationToken);
+        // This call is not awaited because LLamaSharpChatResult accepts an IAsyncEnumerable.
+        var result = this.session.ChatAsync(chat.ToLLamaSharpChatHistory(), settings.ToLLamaSharpInferenceParams(), cancellationToken);
 
         yield return new LLamaSharpChatResult(result);
     }

LLama.SemanticKernel/ExtensionMethods.cs

@@ -1,5 +1,5 @@
-using Microsoft.SemanticKernel.AI.ChatCompletion;
-using Microsoft.SemanticKernel.AI.TextCompletion;
+using LLamaSharp.SemanticKernel.ChatCompletion;
+using Microsoft.SemanticKernel.AI.ChatCompletion;
 
 namespace LLamaSharp.SemanticKernel;
 
@@ -46,27 +46,4 @@ internal static class ExtensionMethods
             MaxTokens = requestSettings.MaxTokens ?? -1
         };
     }
-
-    /// <summary>
-    /// Convert CompleteRequestSettings to LLamaSharp InferenceParams
-    /// </summary>
-    /// <param name="requestSettings"></param>
-    /// <returns></returns>
-    internal static global::LLama.Common.InferenceParams ToLLamaSharpInferenceParams(this CompleteRequestSettings requestSettings)
-    {
-        if (requestSettings is null)
-        {
-            throw new ArgumentNullException(nameof(requestSettings));
-        }
-
-        return new global::LLama.Common.InferenceParams
-        {
-            Temperature = (float)requestSettings.Temperature,
-            TopP = (float)requestSettings.TopP,
-            PresencePenalty = (float)requestSettings.PresencePenalty,
-            FrequencyPenalty = (float)requestSettings.FrequencyPenalty,
-            AntiPrompts = requestSettings.StopSequences,
-            MaxTokens = requestSettings.MaxTokens ?? -1
-        };
-    }
 }

LLama.SemanticKernel/LLamaSharp.SemanticKernel.csproj

@@ -10,7 +10,7 @@
 		<ImplicitUsings>enable</ImplicitUsings>
 		<Nullable>enable</Nullable>
 
-    <Version>0.5.0</Version>
+    <Version>0.6.2-beta1</Version>
     <Authors>Tim Miller</Authors>
     <Company>SciSharp STACK</Company>
     <GeneratePackageOnBuild>true</GeneratePackageOnBuild>
@@ -33,7 +33,7 @@
 	</PropertyGroup>
 
 	<ItemGroup>
-	  <PackageReference Include="Microsoft.SemanticKernel.Abstractions" Version="0.24.230911.2-preview" />
+	  <PackageReference Include="Microsoft.SemanticKernel.Abstractions" Version="1.0.0-beta1" />
 	</ItemGroup>
 
   <ItemGroup Condition="'$(TargetFramework)' == 'netstandard2.0'">

LLama.SemanticKernel/TextCompletion/LLamaSharpTextCompletion.cs

@@ -1,5 +1,6 @@
-using LLama;
-using LLama.Abstractions;
+using LLama.Abstractions;
+using LLamaSharp.SemanticKernel.ChatCompletion;
+using Microsoft.SemanticKernel.AI;
 using Microsoft.SemanticKernel.AI.TextCompletion;
 
 namespace LLamaSharp.SemanticKernel.TextCompletion;
@@ -13,15 +14,17 @@ public sealed class LLamaSharpTextCompletion : ITextCompletion
         this.executor = executor;
     }
 
-    public async Task<IReadOnlyList<ITextResult>> GetCompletionsAsync(string text, CompleteRequestSettings requestSettings, CancellationToken cancellationToken = default)
+    public async Task<IReadOnlyList<ITextResult>> GetCompletionsAsync(string text, AIRequestSettings? requestSettings, CancellationToken cancellationToken = default)
     {
-        var result = executor.InferAsync(text, requestSettings.ToLLamaSharpInferenceParams(), cancellationToken);
+        var settings = (ChatRequestSettings)requestSettings;
+        var result = executor.InferAsync(text, settings?.ToLLamaSharpInferenceParams(), cancellationToken);
         return await Task.FromResult(new List<ITextResult> { new LLamaTextResult(result) }.AsReadOnly()).ConfigureAwait(false);
     }
 
-    public async IAsyncEnumerable<ITextStreamingResult> GetStreamingCompletionsAsync(string text, CompleteRequestSettings requestSettings, CancellationToken cancellationToken = default)
+    public async IAsyncEnumerable<ITextStreamingResult> GetStreamingCompletionsAsync(string text, AIRequestSettings? requestSettings, CancellationToken cancellationToken = default)
     {
-        var result = executor.InferAsync(text, requestSettings.ToLLamaSharpInferenceParams(), cancellationToken);
+        var settings = (ChatRequestSettings)requestSettings;
+        var result = executor.InferAsync(text, settings?.ToLLamaSharpInferenceParams(), cancellationToken);
         yield return new LLamaTextResult(result);
     }
 }