refactor: replace Kokoro-82M with suno/bark for TTS, update HF cache path, and add model warmup on startup.

backend/tests/play_audio.py

@@ -0,0 +1,37 @@
+import asyncio
+import base64
+import os
+import sys
+
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '../..')))
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))
+
+from backend.tts_asr import _tts_sync_with_retry
+
+async def play_audio():
+    print("生成测试音频中，请稍候...")
+    test_text = "这是一段用以测试新语音模型音质的中文合成音频。"
+    
+    try:
+        audio_bytes, sr = await _tts_sync_with_retry(test_text, rate=1.0)
+        
+        # 保存到本地文件
+        wav_path = os.path.join(os.path.dirname(__file__), "test_audio.wav")
+        with open(wav_path, "wb") as f:
+            f.write(audio_bytes)
+            
+        print(f"音频已生成并保存到: {wav_path}")
+        print("正在尝试在 macOS 上播放...")
+        
+        # Mac OS 的播放命令
+        os.system(f"afplay '{wav_path}'")
+        
+        print("播放完成。")
+            
+    except Exception as e:
+        import traceback
+        traceback.print_exc()
+        print(f"音频生成失败: {str(e)}")
+
+if __name__ == "__main__":
+    asyncio.run(play_audio())

backend/tests/test_tts_asr_unit.py

@@ -16,8 +16,9 @@ from unittest.mock import Mock, MagicMock, patch
 import tempfile
 import numpy as np
 
-# 确保可以导入tts_asr模块
-sys.path.insert(0, os.path.join(os.path.dirname(__file__), '..'))
+# 确保可以导入backend和tts_asr模块
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '../..')))
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))
 
 
 class TestAppleSiliconDetection(unittest.TestCase):
@@ -174,7 +175,7 @@ class TestModelSizeSelection(unittest.TestCase):
             import backend.tts_asr as tts_asr_module
             importlib.reload(tts_asr_module)
             
-            from backend.tts_asr import _get_recommended_model_size
+            from backend.tts_asr import _get_recommended_model_size, WHISPER_MODEL_SIZES
             # 应该回退到推荐大小而不崩溃
             size = _get_recommended_model_size()
             self.assertIn(size, WHISPER_MODEL_SIZES.keys())
@@ -299,8 +300,8 @@ class TestModelCacheCheck(unittest.TestCase):
         """测试离线模式下缺失模型的处理"""
         from backend.tts_asr import _check_model_cached
         
-        # 模拟transformers缓存路径
-        with patch('transformers.file_utils.default_cache_path', '/nonexistent/path'):
+        # 模拟缓存路径
+        with patch('huggingface_hub.constants.HF_HUB_CACHE', '/nonexistent/path'):
             result = _check_model_cached('nonexistent/model')
             # 应该返回False（模型未缓存）
             self.assertFalse(result)

backend/tests/verify_cross.py

@@ -0,0 +1,56 @@
+import asyncio
+import base64
+import os
+import sys
+
+# 确保能找到backend模块
+sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '../..')))
+
+from backend.tts_asr import _tts_sync_with_retry, _load_asr_pipeline_with_retry, _get_asr_pipeline
+
+async def verify_tts_asr_cross():
+    print("====================================")
+    print(" 开始严格交叉验证: TTS 生成 -> ASR 解析")
+    print("====================================")
+    
+    test_text = "苹果设备支持离线大模型运算"
+    print(f"\n[1] 正在调用 TTS 引擎 (suno/bark-small)...")
+    print(f"目标文本: '{test_text}'")
+    
+    try:
+        # TTS生成
+        audio_bytes, sr = await _tts_sync_with_retry(test_text, rate=1.0)
+        print(f"-> TTS 成功生成音频数据，大小: {len(audio_bytes)} Bytes, 采样率: {sr}Hz")
+    except Exception as e:
+        print(f"-> TTS 失败: {str(e)}")
+        sys.exit(1)
+        
+    print("\n[2] 正在调用 ASR 引擎 (Whisper)...")
+    try:
+        loaded = await _load_asr_pipeline_with_retry()
+        if not loaded:
+            print("-> ASR 核心加载失败！")
+            sys.exit(1)
+            
+        print("-> ASR 加载成功，开始解析音频...")
+        
+        # 将生成的wav bytes传递给ASR进行语音识别
+        asr_pipeline = _get_asr_pipeline()
+        result = asr_pipeline(audio_bytes, generate_kwargs={"task": "transcribe"})
+        recognized_text = result.get('text', '')
+        print(f"-> ASR 识别结果: '{recognized_text.strip()}'")
+        
+        if len(recognized_text.strip()) > 0:
+            print("\n结论: ✅ 验证成功！TTS和ASR模块功能链路闭环完成。")
+        else:
+            print("\n结论: ❌ ASR输出为空字符，闭环失败。")
+            sys.exit(1)
+            
+    except Exception as e:
+        import traceback
+        traceback.print_exc()
+        print(f"-> ASR 分析阶段失败: {str(e)}")
+        sys.exit(1)
+
+if __name__ == "__main__":
+    asyncio.run(verify_tts_asr_cross())