Add local SenseVoice recognition service

01.VRCLS.Core/Interfaces/IModelPathResolver.cs

@@ -16,5 +16,10 @@ public interface IModelPathResolver
         /// Resolve the Sherpa model directory for local recognition, if available.
         /// </summary>
         string? ResolveSherpaModelDir();
+
+        /// <summary>
+        /// Resolve a complete SenseVoice model directory containing model.onnx and tokens.txt.
+        /// </summary>
+        string? ResolveSenseVoiceModelDir();
     }
 }

01.VRCLS.Core/Models/ServiceErrorCode.cs

@@ -9,6 +9,7 @@ public enum ServiceErrorCode
         AudioProcessingRepeatedFailures,
         RecognitionAuthRequired,
         RecognitionAuthExpired,
+        RecognitionFailed,
         TtsAuthRequired,
         TtsRateLimit,
         TtsAuthExpired,

02.VRCLS.Services/Audio/AudioPipelineFactory.cs

@@ -52,6 +52,9 @@ public AudioPipelineService CreatePipeline(string source, bool isLoopback)
         {
             _logger.Information("Creating audio pipeline. Source: {Source}, Loopback: {IsLoopback}", source, isLoopback);
 
+            AudioCaptureService? capture = null;
+            IVadService? vad = null;
+            IRecognitionService? recognition = null;
             try
             {
                 var useVad = !IsMicStreamingLocalMode(source);
@@ -77,9 +80,8 @@ public AudioPipelineService CreatePipeline(string source, bool isLoopback)
                 }
 
                 _logger.Debug("Creating audio capture service.");
-                var capture = new AudioCaptureService();
+                capture = new AudioCaptureService();
 
-                IVadService? vad = null;
                 if (useVad)
                 {
                     _logger.Debug("Creating VAD service.");
@@ -104,7 +106,7 @@ public AudioPipelineService CreatePipeline(string source, bool isLoopback)
                 }
 
                 _logger.Debug("Creating recognition service.");
-                var recognition = CreateRecognitionService(source, useVad);
+                recognition = CreateRecognitionService(source, useVad);
 
                 _logger.Debug("Creating audio pipeline service.");
                 var pipeline = new AudioPipelineService(
@@ -122,61 +124,83 @@ public AudioPipelineService CreatePipeline(string source, bool isLoopback)
             catch (Exception ex)
             {
                 _logger.Error(ex, "Failed to create audio pipeline. Source: {Source}", source);
+                DisposeFailedPipelineComponent(recognition, "recognition");
+                DisposeFailedPipelineComponent(vad, "VAD");
+                DisposeFailedPipelineComponent(capture, "capture");
                 throw;
             }
         }
 
-        private bool IsMicStreamingLocalMode(string source)
+        private static void DisposeFailedPipelineComponent(object? component, string componentName)
         {
-            if (!string.Equals(source, "Mic", StringComparison.OrdinalIgnoreCase))
+            if (component is not IDisposable disposable)
             {
-                return false;
+                return;
             }
 
-            if (_config.RecognitionBackendModeValue != RecognitionBackendMode.Local)
+            try
             {
-                return false;
+                disposable.Dispose();
             }
-
-            if (_config.LocalRecognitionModelValue == LocalRecognitionModel.Sherpa)
+            catch (Exception ex)
             {
-                return true;
+                _logger.Warning(ex, "Failed to dispose {Component} after pipeline creation failure.", componentName);
             }
+        }
 
-            // If a Sherpa streaming model exists, treat it as streaming even when the model type is not set.
-            var sherpaDir = _modelPathResolver.ResolveSherpaModelDir();
-            return !string.IsNullOrWhiteSpace(sherpaDir);
+        private bool IsMicStreamingLocalMode(string source)
+        {
+            return RecognitionRouteSelector.Select(
+                source,
+                _config.RecognitionBackendModeValue,
+                _config.LocalRecognitionModelValue) == RecognitionRoute.LocalSherpa;
         }
 
         private IRecognitionService CreateRecognitionService(string source, bool useVad)
         {
             var mode = _config.RecognitionBackendModeValue;
             _logger.Debug("Recognition service mode: {Mode}", mode);
 
-            if (string.Equals(source, "Desktop", StringComparison.OrdinalIgnoreCase))
+            var route = RecognitionRouteSelector.Select(
+                source,
+                mode,
+                _config.LocalRecognitionModelValue);
+
+            if (route == RecognitionRoute.Remote)
             {
-                _logger.Information("Desktop source uses remote recognition service only.");
+                _logger.Information("Using remote FastAPI recognition service. Source: {Source}", source);
                 return new FastApiRecognitionService(_config, source, _backendAuthService, _httpClientFactory);
             }
 
-            if (mode == RecognitionBackendMode.Local)
+            if (route == RecognitionRoute.LocalSenseVoice)
             {
-                _logger.Debug("Attempting to use local recognition service for source: {Source}", source);
-
-                var sherpaDir = _modelPathResolver.ResolveSherpaModelDir();
-                if (!string.IsNullOrWhiteSpace(sherpaDir))
+                var senseVoiceDir = _modelPathResolver.ResolveSenseVoiceModelDir();
+                if (string.IsNullOrWhiteSpace(senseVoiceDir))
                 {
-                    _logger.Information("Using local Sherpa recognition service. Source: {Source}, ModelDir: {SherpaDir}", source, sherpaDir);
-                    var endpointSilence = Math.Clamp(_config.LocalStreamingEndpointSilenceSeconds, 0.8f, 3.0f);
-                    return new SherpaAsrService(sherpaDir, source, useVad, endpointSilence);
+                    throw new InvalidOperationException(
+                        "Local SenseVoice model is incomplete. Configure a directory containing model.onnx and tokens.txt.");
                 }
 
-                _logger.Warning("Sherpa model directory not found; falling back to remote recognition.");
+                _logger.Information(
+                    "Using local SenseVoice recognition service. Source: {Source}, ModelDir: {SenseVoiceDir}",
+                    source,
+                    senseVoiceDir);
+                return new SenseVoiceAsrService(senseVoiceDir, source);
+            }
+
+            var sherpaDir = _modelPathResolver.ResolveSherpaModelDir();
+            if (string.IsNullOrWhiteSpace(sherpaDir))
+            {
+                throw new InvalidOperationException(
+                    "Local Sherpa model is incomplete or missing. Check the local model directory.");
             }
 
-            // Default to remote API
-            _logger.Information("Using remote FastAPI recognition service.");
-            return new FastApiRecognitionService(_config, source, _backendAuthService, _httpClientFactory);
+            _logger.Information(
+                "Using local Sherpa recognition service. Source: {Source}, ModelDir: {SherpaDir}",
+                source,
+                sherpaDir);
+            var endpointSilence = Math.Clamp(_config.LocalStreamingEndpointSilenceSeconds, 0.8f, 3.0f);
+            return new SherpaAsrService(sherpaDir, source, useVad, endpointSilence);
         }
 
         private void ValidateOfficialVadChecksum(string vadPath)

02.VRCLS.Services/Audio/ModelPathResolver.cs

@@ -1,6 +1,7 @@
 using System;
 using System.Collections.Generic;
 using System.IO;
+using System.Linq;
 using VRCLS.Core.Interfaces;
 using VRCLS.Core.Models;
 using Serilog;
@@ -151,5 +152,76 @@ private string ResolveRootDir(string baseDir)
             _logger.Warning("未找到Sherpa模型目录");
             return null;
         }
+
+        public string? ResolveSenseVoiceModelDir()
+        {
+            _logger.Debug("开始解析SenseVoice模型目录");
+
+            var baseDir = AppDomain.CurrentDomain.BaseDirectory;
+            var rootDir = ResolveRootDir(baseDir);
+            var preModelsRoot = GetPreModelsRoot(baseDir, rootDir);
+            var candidates = new List<string>();
+
+            if (!string.IsNullOrWhiteSpace(_config.SenseVoiceModelDir))
+            {
+                candidates.Add(_config.SenseVoiceModelDir);
+            }
+
+            candidates.Add(Path.Combine(preModelsRoot, "sensevoice-small"));
+            candidates.Add(Path.Combine(baseDir, "models", "sensevoice-small"));
+            candidates.Add(Path.Combine(rootDir, "models", "sensevoice-small"));
+
+            foreach (var candidate in candidates)
+            {
+                if (IsCompleteSenseVoiceModel(candidate))
+                {
+                    _logger.Information("找到SenseVoice模型 - 路径: {Path}", candidate);
+                    return candidate;
+                }
+            }
+
+            foreach (var searchRoot in new[]
+            {
+                preModelsRoot,
+                Path.Combine(baseDir, "models"),
+                Path.Combine(rootDir, "models")
+            }.Distinct(StringComparer.OrdinalIgnoreCase))
+            {
+                if (!Directory.Exists(searchRoot))
+                {
+                    continue;
+                }
+
+                try
+                {
+                    foreach (var modelPath in Directory.EnumerateFiles(
+                        searchRoot,
+                        "model.onnx",
+                        SearchOption.AllDirectories))
+                    {
+                        var modelDir = Path.GetDirectoryName(modelPath);
+                        if (IsCompleteSenseVoiceModel(modelDir))
+                        {
+                            _logger.Information("找到SenseVoice模型 - 路径: {Path}", modelDir);
+                            return modelDir;
+                        }
+                    }
+                }
+                catch (Exception ex)
+                {
+                    _logger.Debug(ex, "搜索SenseVoice模型目录失败: {Path}", searchRoot);
+                }
+            }
+
+            _logger.Warning("未找到完整SenseVoice模型目录");
+            return null;
+        }
+
+        private static bool IsCompleteSenseVoiceModel(string? modelDir)
+        {
+            return !string.IsNullOrWhiteSpace(modelDir)
+                && File.Exists(Path.Combine(modelDir, "model.onnx"))
+                && File.Exists(Path.Combine(modelDir, "tokens.txt"));
+        }
     }
 }

02.VRCLS.Services/Audio/RecognitionRouteSelector.cs

@@ -0,0 +1,31 @@
+using System;
+using VRCLS.Core.Models;
+
+namespace VRCLS.Services.Audio
+{
+    internal enum RecognitionRoute
+    {
+        Remote,
+        LocalSherpa,
+        LocalSenseVoice
+    }
+
+    internal static class RecognitionRouteSelector
+    {
+        public static RecognitionRoute Select(
+            string source,
+            RecognitionBackendMode mode,
+            LocalRecognitionModel localModel)
+        {
+            if (!string.Equals(source, "Mic", StringComparison.OrdinalIgnoreCase)
+                || mode != RecognitionBackendMode.Local)
+            {
+                return RecognitionRoute.Remote;
+            }
+
+            return localModel == LocalRecognitionModel.SenseVoice
+                ? RecognitionRoute.LocalSenseVoice
+                : RecognitionRoute.LocalSherpa;
+        }
+    }
+}

02.VRCLS.Services/Recognition/IOfflineSenseVoiceRecognizer.cs

@@ -0,0 +1,9 @@
+using System;
+
+namespace VRCLS.Services.Recognition
+{
+    internal interface IOfflineSenseVoiceRecognizer : IDisposable
+    {
+        string Decode(float[] audioData);
+    }
+}

02.VRCLS.Services/Recognition/SenseVoiceAsrService.Sherpa.cs

@@ -0,0 +1,10 @@
+namespace VRCLS.Services.Recognition
+{
+    public sealed partial class SenseVoiceAsrService
+    {
+        public SenseVoiceAsrService(string modelPath, string source = "Mic")
+            : this(new SherpaOfflineSenseVoiceRecognizer(modelPath), source)
+        {
+        }
+    }
+}