重构语音能力配置与逻辑，统一音频输入输出开关并优化语音回复判断

tests/test_agent_image_support.py

@@ -226,14 +226,17 @@ class AgentImageSupportTest(unittest.TestCase):
         ), patch.object(chain.messagehelper, "put"), patch.object(
             chain.messageoper, "add"
         ), patch.object(chain, "_handle_ai_message") as handle_ai_message:
-            chain.handle_message(
-                channel=MessageChannel.Telegram,
-                source="telegram-test",
-                userid="10001",
-                username="tester",
-                text="",
-                audio_refs=["tg://voice_file_id/voice-1"],
-            )
+            with patch.object(settings, "AI_AGENT_ENABLE", True), patch.object(
+                settings, "AI_AGENT_GLOBAL", False
+            ):
+                chain.handle_message(
+                    channel=MessageChannel.Telegram,
+                    source="telegram-test",
+                    userid="10001",
+                    username="tester",
+                    text="",
+                    audio_refs=["tg://voice_file_id/voice-1"],
+                )
 
         handle_ai_message.assert_called_once()
         self.assertEqual(handle_ai_message.call_args.kwargs["text"], "帮我推荐一部电影")
@@ -319,7 +322,7 @@ class AgentImageSupportTest(unittest.TestCase):
             ],
         )
 
-    def test_agent_send_agent_message_does_not_auto_convert_to_voice(self):
+    def test_agent_send_agent_message_auto_converts_to_voice_when_supported(self):
         agent = MoviePilotAgent(
             session_id="session-1",
             user_id="user-1",
@@ -330,6 +333,14 @@ class AgentImageSupportTest(unittest.TestCase):
         agent.reply_with_voice = True
 
         with patch.object(
+            VoiceHelper,
+            "resolve_reply_mode",
+            return_value=VoiceHelper.REPLY_MODE_NATIVE,
+        ), patch.object(
+            VoiceHelper, "is_available", return_value=True
+        ), patch.object(
+            VoiceHelper, "synthesize_speech", return_value=Path("/tmp/reply.opus")
+        ), patch.object(
             AgentChain, "async_post_message", new_callable=AsyncMock
         ) as async_post_message:
             import asyncio
@@ -337,7 +348,7 @@ class AgentImageSupportTest(unittest.TestCase):
             asyncio.run(agent.send_agent_message("这是语音回复"))
 
         notification = async_post_message.await_args.args[0]
-        self.assertIsNone(notification.voice_path)
+        self.assertEqual(notification.voice_path, "/tmp/reply.opus")
         self.assertEqual(notification.text, "这是语音回复")
 
     def test_agent_process_wraps_request_as_structured_json(self):

tests/test_voice_helper.py

@@ -13,10 +13,8 @@ class VoiceHelperTest(unittest.TestCase):
     def test_registered_providers_contains_openai(self):
         self.assertIn("openai", VoiceHelper.get_registered_providers())
 
-    def test_get_provider_falls_back_to_global_provider(self):
-        with patch.object(settings, "AI_VOICE_PROVIDER", "openai"), patch.object(
-            settings, "AI_VOICE_STT_PROVIDER", None
-        ):
+    def test_get_provider_uses_single_audio_provider_setting(self):
+        with patch.object(settings, "AI_VOICE_PROVIDER", "openai"):
             provider = VoiceHelper.get_provider("stt")
 
         self.assertIsInstance(provider, OpenAIVoiceProvider)
@@ -26,15 +24,29 @@ class VoiceHelperTest(unittest.TestCase):
         provider.is_available_for_stt.return_value = True
         provider.is_available_for_tts.return_value = False
 
-        with patch.object(VoiceHelper, "get_provider", return_value=provider):
+        with patch.object(
+            settings, "LLM_SUPPORT_AUDIO_INPUT_OUTPUT", True
+        ), patch.object(VoiceHelper, "get_provider", return_value=provider):
             self.assertTrue(VoiceHelper.is_available("stt"))
             self.assertFalse(VoiceHelper.is_available("tts"))
 
+    def test_is_available_returns_false_when_audio_switch_is_disabled(self):
+        provider = Mock()
+        provider.is_available_for_stt.return_value = True
+
+        with patch.object(
+            settings, "LLM_SUPPORT_AUDIO_INPUT_OUTPUT", False
+        ), patch.object(VoiceHelper, "get_provider", return_value=provider):
+            self.assertFalse(VoiceHelper.is_available("stt"))
+            self.assertFalse(VoiceHelper.is_available())
+
     def test_transcribe_bytes_routes_to_stt_provider(self):
         provider = Mock()
         provider.transcribe_bytes.return_value = "你好"
 
-        with patch.object(VoiceHelper, "get_provider", return_value=provider):
+        with patch.object(
+            settings, "LLM_SUPPORT_AUDIO_INPUT_OUTPUT", True
+        ), patch.object(VoiceHelper, "get_provider", return_value=provider):
             result = VoiceHelper.transcribe_bytes(b"audio")
 
         self.assertEqual(result, "你好")
@@ -44,7 +56,9 @@ class VoiceHelperTest(unittest.TestCase):
         provider = Mock()
         provider.synthesize_speech.return_value = "/tmp/reply.opus"
 
-        with patch.object(VoiceHelper, "get_provider", return_value=provider):
+        with patch.object(
+            settings, "LLM_SUPPORT_AUDIO_INPUT_OUTPUT", True
+        ), patch.object(VoiceHelper, "get_provider", return_value=provider):
             result = VoiceHelper.synthesize_speech("你好")
 
         self.assertEqual(result, "/tmp/reply.opus")