debug: 增加CosyVoice请求/响应详细日志，用于排查418错误

packages/application/cosyvoice_service.py

@@ -70,21 +70,21 @@ class CosyVoiceService:
     - 音色克隆: POST /services/audio/tts/customization (model=voice-enrollment)
       - action=create_voice: 创建克隆音色，返回 voice_id（状态 DEPLOYING）
       - action=query_voice: 查询音色状态（DEPLOYING / OK / UNDEPLOYED）
-    - 语音合成: POST /services/audio/tts/SpeechSynthesizer (model=cosyvoice-v3.5-plus)
+    - 语音合成: POST /services/audio/tts/SpeechSynthesizer (model=cosyvoice-v3-flash)
       - 非流式: 同步返回音频 URL
 
     使用示例:
         service = CosyVoiceService(
             api_key="your-api-key",
             base_url="https://dashscope.aliyuncs.com/api/v1",
-            model="cosyvoice-v3.5-plus",
+            model="cosyvoice-v3-flash",
         )
 
         # 音色克隆
         result = service.clone_voice(audio_url="https://example.com/audio.mp3")
 
         # 语音合成
-        result = service.synthesize_speech(text="你好世界", voice_id="longxiaochun")
+        result = service.synthesize_speech(text="你好世界", voice_id="longxiaochun_v3")
     """
 
     # 音色状态轮询配置
@@ -589,6 +589,20 @@ class CosyVoiceService:
             "Content-Type": "application/json",
         }
 
+        # DEBUG: 打印完整请求信息，用于排查418错误
+        import json as json_lib
+        safe_headers = {k: v for k, v in headers.items()}
+        if "Authorization" in safe_headers:
+            token = safe_headers["Authorization"]
+            if len(token) > 20:
+                safe_headers["Authorization"] = token[:13] + "..." + token[-4:]
+        logger.info(
+            "[CosyVoice Debug] 请求详情: "
+            "method=%s, url=%s, headers=%s, body=%s",
+            method, url, safe_headers,
+            json_lib.dumps(json, ensure_ascii=False) if json else "None",
+        )
+
         last_error: Optional[Exception] = None
 
         for attempt in range(self.MAX_RETRIES):
@@ -601,6 +615,14 @@ class CosyVoiceService:
                     timeout=timeout,
                 )
 
+                # DEBUG: 打印响应状态和完整响应体
+                logger.info(
+                    "[CosyVoice Debug] 响应详情: "
+                    "status=%d, body=%s",
+                    response.status_code,
+                    response.text[:2000],  # 最多2000字符，避免日志过大
+                )
+
                 # 处理响应
                 if response.status_code == 200:
                     return response.json()

packages/domain/preset_voices.py

@@ -16,7 +16,7 @@ class PresetVoice:
     """预置音色定义。
 
     Attributes:
-        voice_id: CosyVoice 模型音色名（如 longxiaochun）
+        voice_id: CosyVoice 模型音色名（如 longxiaochun_v3）
         name: 中文展示名
         description: 音色描述
         gender: 性别（male/female）
@@ -49,7 +49,7 @@ class PresetVoice:
 # 预置音色列表（阿里云 CosyVoice 真实可用音色）
 PRESET_VOICES: list[PresetVoice] = [
     PresetVoice(
-        voice_id="longxiaochun",
+        voice_id="longxiaochun_v3",
         name="龙小淳",
         description="温柔女声，适合情感类内容",
         gender="female",
@@ -57,7 +57,7 @@ PRESET_VOICES: list[PresetVoice] = [
         tags=["温柔", "女声", "情感"],
     ),
     PresetVoice(
-        voice_id="longxiaoxia",
+        voice_id="longxiaoxia_v3",
         name="龙小夏",
         description="知性女声，适合新闻播报",
         gender="female",
@@ -65,7 +65,7 @@ PRESET_VOICES: list[PresetVoice] = [
         tags=["知性", "女声", "播报"],
     ),
     PresetVoice(
-        voice_id="longxiaochen",
+        voice_id="longxiaochen_v3",
         name="龙小晨",
         description="磁性男声，适合有声书",
         gender="male",
@@ -73,7 +73,7 @@ PRESET_VOICES: list[PresetVoice] = [
         tags=["磁性", "男声", "有声书"],
     ),
     PresetVoice(
-        voice_id="longyue",
+        voice_id="longyue_v3",
         name="龙悦",
         description="甜美女声，适合广告配音",
         gender="female",
@@ -81,7 +81,7 @@ PRESET_VOICES: list[PresetVoice] = [
         tags=["甜美", "女声", "广告"],
     ),
     PresetVoice(
-        voice_id="longshu",
+        voice_id="longshu_v3",
         name="龙书",
         description="沉稳男声，适合教育讲解",
         gender="male",
@@ -89,7 +89,7 @@ PRESET_VOICES: list[PresetVoice] = [
         tags=["沉稳", "男声", "教育"],
     ),
     PresetVoice(
-        voice_id="longjing",
+        voice_id="longjing_v3",
         name="龙静",
         description="优雅女声，适合纪录片解说",
         gender="female",
@@ -97,7 +97,7 @@ PRESET_VOICES: list[PresetVoice] = [
         tags=["优雅", "女声", "纪录片"],
     ),
     PresetVoice(
-        voice_id="longbo",
+        voice_id="longbo_v3",
         name="龙博",
         description="浑厚男声，适合科技类内容",
         gender="male",
@@ -105,7 +105,7 @@ PRESET_VOICES: list[PresetVoice] = [
         tags=["浑厚", "男声", "科技"],
     ),
     PresetVoice(
-        voice_id="longtian",
+        voice_id="longtian_v3",
         name="龙甜",
         description="活泼女声，适合短视频配音",
         gender="female",

packages/shared/config.py

@@ -32,8 +32,8 @@ class SharedSettings(BaseSettings):
     # CosyVoice (阿里云百炼语音合成)
     cosyvoice_api_key: str = ""
     cosyvoice_base_url: str = "https://dashscope.aliyuncs.com/api/v1"
-    cosyvoice_model: str = "cosyvoice-v3.5-plus"
-    cosyvoice_voice: str = "longxiaochun"  # 默认音色
+    cosyvoice_model: str = "cosyvoice-v3-flash"
+    cosyvoice_voice: str = "longxiaochun_v3"  # 默认音色（v3 系列系统音色带 _v3 后缀）
     cosyvoice_sample_rate: int = 22050
     cosyvoice_format: str = "mp3"  # 输出格式：mp3/wav/pcm
     # 音色克隆模型名（固定为 voice-enrollment）

tests/unit/test_cosyvoice_service.py

@@ -23,7 +23,7 @@ def _make_service(
     *,
     api_key: str = "test-api-key",
     base_url: str = "https://dashscope.aliyuncs.com/api/v1",
-    model: str = "cosyvoice-v3.5-plus",
+    model: str = "cosyvoice-v3-flash",
     clone_model: str = "voice-enrollment",
     http_client: httpx.Client | None = None,
     audio_url_signer=None,
@@ -78,7 +78,7 @@ class TestSubmitCloneTask:
             200,
             {
                 "output": {
-                    "voice_id": "cosyvoice-v3.5-plus-clone-abc123",
+                    "voice_id": "cosyvoice-v3-flash-clone-abc123",
                     "status": "DEPLOYING",
                 },
                 "usage": {"count": 1},
@@ -92,7 +92,7 @@ class TestSubmitCloneTask:
             voice_name="myvoice",
         )
 
-        assert result["voice_id"] == "cosyvoice-v3.5-plus-clone-abc123"
+        assert result["voice_id"] == "cosyvoice-v3-flash-clone-abc123"
         assert result["status"] == "DEPLOYING"
         assert result["request_id"] == "req-001"
 
@@ -104,7 +104,7 @@ class TestSubmitCloneTask:
         payload = call_args.kwargs["json"]
         assert payload["model"] == "voice-enrollment"
         assert payload["input"]["action"] == "create_voice"
-        assert payload["input"]["target_model"] == "cosyvoice-v3.5-plus"
+        assert payload["input"]["target_model"] == "cosyvoice-v3-flash"
         assert payload["input"]["prefix"] == "myvoice"
         assert payload["input"]["url"] == "https://example.com/audio.wav"
         assert payload["input"]["language_hints"] == ["zh"]
@@ -229,7 +229,7 @@ class TestQueryVoiceStatus:
             {
                 "output": {
                     "status": "DEPLOYING",
-                    "target_model": "cosyvoice-v3.5-plus",
+                    "target_model": "cosyvoice-v3-flash",
                     "gmt_create": "2026-01-01T00:00:00Z",
                     "gmt_modified": "2026-01-01T00:01:00Z",
                     "resource_link": "https://...",
@@ -242,7 +242,7 @@ class TestQueryVoiceStatus:
         result = service.query_voice_status("voice-123")
 
         assert result["status"] == "DEPLOYING"
-        assert result["target_model"] == "cosyvoice-v3.5-plus"
+        assert result["target_model"] == "cosyvoice-v3-flash"
 
         # 验证请求
         payload = mock_client.request.call_args.kwargs["json"]
@@ -253,7 +253,7 @@ class TestQueryVoiceStatus:
     def test_query_ok_status(self) -> None:
         mock_client = MagicMock()
         mock_client.request.return_value = _mock_response(
-            200, {"output": {"status": "OK", "target_model": "cosyvoice-v3.5-plus"}}
+            200, {"output": {"status": "OK", "target_model": "cosyvoice-v3-flash"}}
         )
 
         service = _make_service(http_client=mock_client)
@@ -278,7 +278,7 @@ class TestPollCloneTask:
     def test_poll_ok_on_first_check(self) -> None:
         mock_client = MagicMock()
         mock_client.request.return_value = _mock_response(
-            200, {"output": {"status": "OK", "target_model": "cosyvoice-v3.5-plus"}}
+            200, {"output": {"status": "OK", "target_model": "cosyvoice-v3-flash"}}
         )
 
         service = _make_service(http_client=mock_client)
@@ -397,7 +397,7 @@ class TestSynthesizeSpeech:
 
         service = _make_service(http_client=mock_client)
         result = service.synthesize_speech(
-            text="你好世界", voice_id="longxiaochun"
+            text="你好世界", voice_id="longxiaochun_v3"
         )
 
         assert isinstance(result, SynthesizeResult)
@@ -409,9 +409,9 @@ class TestSynthesizeSpeech:
         assert "/services/audio/tts/SpeechSynthesizer" in call_args.kwargs["url"]
 
         payload = call_args.kwargs["json"]
-        assert payload["model"] == "cosyvoice-v3.5-plus"
+        assert payload["model"] == "cosyvoice-v3-flash"
         assert payload["input"]["text"] == "你好世界"
-        assert payload["input"]["voice"] == "longxiaochun"
+        assert payload["input"]["voice"] == "longxiaochun_v3"
         assert payload["input"]["format"] == "mp3"
         assert payload["input"]["sample_rate"] == 22050
         assert payload["input"]["rate"] == 1.0

tests/unit/test_preset_voices.py

@@ -64,7 +64,7 @@ class TestPresetVoice:
     def test_preset_voice_to_dict(self) -> None:
         """序列化。"""
         voice = PresetVoice(
-            voice_id="longxiaochun",
+            voice_id="longxiaochun_v3",
             name="龙小淳",
             description="温柔女声",
             gender="female",
@@ -73,7 +73,7 @@ class TestPresetVoice:
 
         result = voice.to_dict()
 
-        assert result["voice_id"] == "longxiaochun"
+        assert result["voice_id"] == "longxiaochun_v3"
         assert result["name"] == "龙小淳"
         assert result["description"] == "温柔女声"
         assert result["gender"] == "female"
@@ -127,14 +127,14 @@ class TestPresetVoicesConfig:
     def test_cosyvoice_voice_ids(self) -> None:
         """音色 ID 应为 CosyVoice 真实可用的音色名。"""
         expected_ids = {
-            "longxiaochun",
-            "longxiaoxia",
-            "longxiaochen",
-            "longyue",
-            "longshu",
-            "longjing",
-            "longbo",
-            "longtian",
+            "longxiaochun_v3",
+            "longxiaoxia_v3",
+            "longxiaochen_v3",
+            "longyue_v3",
+            "longshu_v3",
+            "longjing_v3",
+            "longbo_v3",
+            "longtian_v3",
         }
         actual_ids = {v.voice_id for v in PRESET_VOICES}
         assert actual_ids == expected_ids
@@ -164,10 +164,10 @@ class TestPresetVoiceHelpers:
 
     def test_get_preset_voice_by_id_found(self) -> None:
         """按 ID 查找存在的音色。"""
-        voice = get_preset_voice_by_id("longxiaochun")
+        voice = get_preset_voice_by_id("longxiaochun_v3")
         assert voice is not None
         assert voice.name == "龙小淳"
-        assert voice.voice_id == "longxiaochun"
+        assert voice.voice_id == "longxiaochun_v3"
 
     def test_get_preset_voice_by_id_not_found(self) -> None:
         """按 ID 查找不存在的音色。"""
@@ -176,9 +176,9 @@ class TestPresetVoiceHelpers:
 
     def test_is_preset_voice_true(self) -> None:
         """判断预置音色返回 True。"""
-        assert is_preset_voice("longxiaochun") is True
-        assert is_preset_voice("longxiaoxia") is True
-        assert is_preset_voice("longbo") is True
+        assert is_preset_voice("longxiaochun_v3") is True
+        assert is_preset_voice("longxiaoxia_v3") is True
+        assert is_preset_voice("longbo_v3") is True
 
     def test_is_preset_voice_false(self) -> None:
         """判断非预置音色返回 False。"""

tests/unit/test_tts_job.py

@@ -15,7 +15,7 @@ class TestTTSJobCreate:
         job = TTSJob.create(
             user_id="user_001",
             input_text="这是一段测试文本",
-            voice_id="longxiaochun",
+            voice_id="longxiaochun_v3",
             voice_model="cosyvoice-v1",
             project_id="project_001",
             voice_clone_profile_id="profile_001",
@@ -26,7 +26,7 @@ class TestTTSJobCreate:
         assert job.id
         assert job.user_id == "user_001"
         assert job.input_text == "这是一段测试文本"
-        assert job.voice_id == "longxiaochun"
+        assert job.voice_id == "longxiaochun_v3"
         assert job.voice_model == "cosyvoice-v1"
         assert job.project_id == "project_001"
         assert job.voice_clone_profile_id == "profile_001"
@@ -291,7 +291,7 @@ class TestTTSJobToDict:
         job = TTSJob.create(
             user_id="user_001",
             input_text="测试文本",
-            voice_id="longxiaochun",
+            voice_id="longxiaochun_v3",
             voice_model="cosyvoice-v1",
             project_id="project_001",
             voice_clone_profile_id="profile_001",
@@ -306,7 +306,7 @@ class TestTTSJobToDict:
         assert result["id"] == job.id
         assert result["user_id"] == "user_001"
         assert result["input_text"] == "测试文本"
-        assert result["voice_id"] == "longxiaochun"
+        assert result["voice_id"] == "longxiaochun_v3"
         assert result["voice_model"] == "cosyvoice-v1"
         assert result["project_id"] == "project_001"
         assert result["voice_clone_profile_id"] == "profile_001"