Merge branch 'main' into 22-agent-knowledge-base

Author: wisedev-pstach

Examples/Examples/Chat/ChatWithTextToSpeechExample.cs

@@ -0,0 +1,28 @@
+using MaIN.Core.Hub;
+using MaIN.Domain.Entities;
+using MaIN.Services.Services.TTSService;
+
+namespace Examples;
+
+public class ChatWithTextToSpeechExample : IExample
+{
+    private const string VoicePath = "<your-path-to-voices>";
+    
+    public async Task Start()
+    {
+        Console.WriteLine("ChatWithTextToSpeech is running! Put on your headphones and press any key.");
+        Console.ReadKey();
+        
+        VoiceService.SetVoicesPath(VoicePath);
+        var voice = VoiceService.GetVoice("af_heart")
+            .MixWith(VoiceService.GetVoice("bf_emma"));
+        
+        await AIHub.Chat().WithModel("gemma2:2b")
+            .WithMessage("Generate a 4 sentence poem.")
+            .Speak(new TextToSpeechParams("kokoro:82m", voice, true))
+            .CompleteAsync(interactive: true);
+
+        Console.WriteLine("Done!");
+        Console.ReadKey();
+    }
+}

Examples/Examples/Program.cs

@@ -71,6 +71,7 @@ static void RegisterExamples(IServiceCollection services)
     services.AddTransient<ChatWithImageGenGeminiExample>();
     services.AddTransient<ChatWithFilesExampleGemini>();
     services.AddTransient<ChatWithReasoningDeepSeekExample>();
+    services.AddTransient<ChatWithTextToSpeechExample>();
     services.AddTransient<ChatExampleGroqCloud>();
 }
 
@@ -163,6 +164,8 @@ public class ExampleRegistry(IServiceProvider serviceProvider)
             ("\u25a0 DeepSeek Chat with reasoning", serviceProvider.GetRequiredService<ChatWithReasoningDeepSeekExample>()),
             ("\u25a0 GroqCloud Chat", serviceProvider.GetRequiredService<ChatExampleGroqCloud>()),
             ("\u25a0 McpClient example", serviceProvider.GetRequiredService<McpExample>()),
+            ("\u25a0 McpAgent example", serviceProvider.GetRequiredService<McpAgentsExample>()),
+            ("\u25a0 Chat with TTS example", serviceProvider.GetRequiredService<ChatWithTextToSpeechExample>()),
             ("\u25a0 McpAgent example", serviceProvider.GetRequiredService<McpAgentsExample>())
         };
     }

Examples/Examples/appsettings.json

@@ -6,10 +6,7 @@
     }
   },
   "MaIN": {
-    "ImageGenUrl": "http://localhost:5003",
-    //    "SqliteSettings": {
-//      "ConnectionString": "Data Source=Main_test.db"
-//    }
+    "ImageGenUrl": "http://localhost:5003"
 },
   "AllowedHosts": "*"
 }

Releases/0.4.0.md

@@ -0,0 +1,3 @@
+# 0.3.1 release
+
+- Added TextToSpeech functionality to using Kokoro TTS model. 

src/MaIN.Core/Hub/Contexts/ChatContext.cs

@@ -3,7 +3,6 @@
 using MaIN.Domain.Models;
 using MaIN.Services;
 using MaIN.Services.Constants;
-using MaIN.Services.Dtos;
 using MaIN.Services.Services.Abstract;
 using MaIN.Services.Services.Models;
 using FileInfo = MaIN.Domain.Entities.FileInfo;
@@ -61,6 +60,14 @@ public ChatContext WithCustomModel(string model, string path, string? mmProject
         return this;
     }
 
+    public ChatContext Speak(TextToSpeechParams textToSpeechParams)
+    {
+        _chat.Visual = false;
+        _chat.TextToSpeechParams = textToSpeechParams;
+        
+        return this;
+    }
+
     public ChatContext WithBackend(BackendType backendType)
     {
         _chat.Backend = backendType;
@@ -151,6 +158,7 @@ public ChatContext WithFiles(List<string> filePaths, bool preProcess = false)
     public ChatContext EnableVisual()
     {
         _chat.Visual = true;
+        _chat.TextToSpeechParams = null;
         return this;
     }
 

src/MaIN.Domain/Configuration/MaINSettings.cs

@@ -16,6 +16,7 @@ public class MaINSettings
     public FileSystemSettings? FileSystemSettings { get; set; }
     public SqliteSettings? SqliteSettings { get; set; }
     public SqlSettings? SqlSettings { get; set; }
+    public string? VoicesPath { get; set; }
 }
 
 public enum BackendType

src/MaIN.Domain/Entities/Chat.cs

@@ -13,6 +13,7 @@ public class Chat
     public bool Visual { get; set; }
     public InferenceParams InterferenceParams { get; set; } = new();
     public MemoryParams MemoryParams { get; set; } = new();
+    public TextToSpeechParams? TextToSpeechParams { get; set; }
     public Dictionary<string, string> Properties { get; init; } = [];
     public List<string> Memory { get; } = [];
     public BackendType? Backend { get; set; }

src/MaIN.Domain/Entities/Message.cs

@@ -20,6 +20,7 @@ public Message()
     public bool Tool { get; init; }
     public DateTime Time { get; set; }
     public byte[]? Image { get; init; }
+    public byte[]? Speech { get; set; }
     public List<FileInfo>? Files { get; set; } 
     public Dictionary<string, string> Properties { get; set; } = [];
 

src/MaIN.Domain/Entities/TextToSpeechParams.cs

@@ -0,0 +1,15 @@
+namespace MaIN.Domain.Entities;
+
+public class TextToSpeechParams
+{
+    public string Model { get; set; }
+    public Voice Voice { get;  set; }
+    public bool Playback { get; set; }
+
+    public TextToSpeechParams(string model, Voice voice, bool playback = false)
+    {
+        Model = model;  
+        Voice = voice;
+        Playback = playback;
+    }
+}

src/MaIN.Domain/Entities/Voice.cs

@@ -0,0 +1,7 @@
+namespace MaIN.Domain.Entities;
+
+public class Voice
+{
+    public required string Name { get; set; }
+    public required float[,,] Features { get; set; }
+}