fix(agent): enable voice replies for supported channels

2026-06-11 18:50:59 +08:00 · 2026-05-26 20:14:56 +08:00
parent db8363fee1
commit 896631d63e
5 changed files with 208 additions and 23 deletions
--- a/tests/test_agent_llm_capability.py
+++ b/tests/test_agent_llm_capability.py
@@ -11,6 +11,8 @@ sys.modules.setdefault("psutil", Mock())
 sys.modules.setdefault("pyquery", Mock())

 from app.core.config import settings
+from app.schemas.message import ChannelCapability, ChannelCapabilityManager
+from app.schemas.types import MessageChannel

 module_path = Path(__file__).resolve().parents[1] / "app" / "agent" / "llm" / "capability.py"
 spec = importlib.util.spec_from_file_location("test_agent_llm_capability_module", module_path)
@@ -157,6 +159,73 @@ class AgentCapabilityManagerTest(unittest.TestCase):
        self.assertEqual(result, Path("/tmp/reply.opus"))
        provider.synthesize_speech.assert_called_once_with(text="你好")

+    def test_native_voice_reply_supports_channels_with_audio_output(self):
+        """校验 Agent 语音回复渠道支持判断覆盖常见渠道写法。"""
+        self.assertTrue(
+            AgentCapabilityManager.supports_native_voice_reply("telegram", None)
+        )
+        self.assertTrue(
+            AgentCapabilityManager.supports_native_voice_reply(
+                MessageChannel.Telegram.value, None
+            )
+        )
+        self.assertTrue(
+            AgentCapabilityManager.supports_native_voice_reply(
+                MessageChannel.Feishu.value, None
+            )
+        )
+        self.assertTrue(
+            AgentCapabilityManager.supports_native_voice_reply("Feishu", None)
+        )
+        self.assertFalse(
+            AgentCapabilityManager.supports_native_voice_reply("Slack", None)
+        )
+
+    def test_native_voice_reply_respects_wechat_mode(self):
+        """校验企业微信只有自建应用模式允许 Agent 语音回复。"""
+        configs = [
+            SimpleNamespace(name="wechat-app", config={"WECHAT_MODE": "app"}),
+            SimpleNamespace(name="wechat-bot", config={"WECHAT_MODE": "bot"}),
+        ]
+
+        with patch(
+            "app.helper.service.ServiceConfigHelper.get_notification_configs",
+            return_value=configs,
+        ):
+            self.assertTrue(
+                AgentCapabilityManager.supports_native_voice_reply(
+                    MessageChannel.Wechat.value, "wechat-app"
+                )
+            )
+            self.assertFalse(
+                AgentCapabilityManager.supports_native_voice_reply(
+                    MessageChannel.Wechat.value, "wechat-bot"
+                )
+            )
+            self.assertFalse(
+                AgentCapabilityManager.supports_native_voice_reply(
+                    MessageChannel.Wechat.value, "missing"
+                )
+            )
+
+    def test_channel_capability_marks_voice_output_channels(self):
+        """校验消息渠道能力显式声明原生语音输出支持。"""
+        for channel in (
+            MessageChannel.Telegram,
+            MessageChannel.Feishu,
+            MessageChannel.Wechat,
+        ):
+            self.assertTrue(
+                ChannelCapabilityManager.supports_capability(
+                    channel, ChannelCapability.AUDIO_OUTPUT
+                )
+            )
+        self.assertFalse(
+            ChannelCapabilityManager.supports_capability(
+                MessageChannel.Slack, ChannelCapability.AUDIO_OUTPUT
+            )
+        )
+
    def test_mimo_tts_uses_chat_completions_audio_payload(self):
        provider = MiMoAudioProvider()
        fake_client = Mock()
--- a/tests/test_agent_tool_streaming.py
+++ b/tests/test_agent_tool_streaming.py
@@ -1,5 +1,6 @@
 import asyncio
 import unittest
+from pathlib import Path
 from unittest.mock import AsyncMock, patch

 import langchain.agents as langchain_agents
@@ -9,6 +10,7 @@ if not hasattr(langchain_agents, "create_agent"):

 from app.agent.callback import StreamingHandler
 from app.agent.tools.base import MoviePilotTool
+from app.agent.tools.impl.send_voice_message import SendVoiceMessageTool
 from app.api.endpoints.openai import _OpenAIStreamingHandler
 from app.core.config import settings
 from app.schemas.message import MessageResponse
@@ -397,6 +399,80 @@ class TestAgentToolStreaming(unittest.TestCase):
        send_tool_message.assert_awaited_once_with("前置内容\n\n⚙️ => run test tool")
        self.assertEqual(buffered_message, "")

+    def test_send_voice_message_uses_native_voice_for_supported_channels(self):
+        """校验支持语音输出的渠道会发送原生语音消息。"""
+
+        async def _run(channel: MessageChannel):
+            """运行指定渠道的语音发送工具。"""
+            tool = SendVoiceMessageTool(session_id="session-1", user_id="10001")
+            tool.set_message_attr(
+                channel=channel.value, source=f"{channel.name.lower()}-main", username="tester"
+            )
+
+            with (
+                patch.object(settings, "LLM_SUPPORT_AUDIO_OUTPUT", True),
+                patch.object(settings, "AUDIO_OUTPUT_INCLUDE_TEXT", True),
+                patch(
+                    "app.agent.tools.impl.send_voice_message.AgentCapabilityManager.is_audio_output_available",
+                    return_value=True,
+                ),
+                patch(
+                    "app.agent.tools.impl.send_voice_message.AgentCapabilityManager.synthesize_speech",
+                    return_value=Path("/tmp/reply.opus"),
+                ) as synthesize_speech,
+                patch(
+                    "app.agent.tools.impl.send_voice_message.ToolChain.async_post_message",
+                    new_callable=AsyncMock,
+                ) as async_post_message,
+            ):
+                result = await tool.run("你好")
+            return result, synthesize_speech, async_post_message
+
+        for channel in (MessageChannel.Telegram, MessageChannel.Feishu):
+            result, synthesize_speech, async_post_message = asyncio.run(_run(channel))
+            notification = async_post_message.await_args.args[0]
+
+            self.assertEqual(result, "语音回复已发送")
+            synthesize_speech.assert_called_once_with("你好")
+            self.assertEqual(notification.channel, channel)
+            self.assertEqual(notification.voice_path, "/tmp/reply.opus")
+            self.assertEqual(notification.voice_caption, "你好")
+
+    def test_send_voice_message_falls_back_for_unsupported_channels(self):
+        """校验不支持语音输出的渠道继续回退为文字消息。"""
+
+        async def _run():
+            """运行不支持语音输出渠道的语音发送工具。"""
+            tool = SendVoiceMessageTool(session_id="session-1", user_id="10001")
+            tool.set_message_attr(
+                channel=MessageChannel.Slack.value, source="slack-main", username="tester"
+            )
+
+            with (
+                patch.object(settings, "LLM_SUPPORT_AUDIO_OUTPUT", True),
+                patch(
+                    "app.agent.tools.impl.send_voice_message.AgentCapabilityManager.is_audio_output_available",
+                    return_value=True,
+                ),
+                patch(
+                    "app.agent.tools.impl.send_voice_message.AgentCapabilityManager.synthesize_speech"
+                ) as synthesize_speech,
+                patch(
+                    "app.agent.tools.impl.send_voice_message.ToolChain.async_post_message",
+                    new_callable=AsyncMock,
+                ) as async_post_message,
+            ):
+                result = await tool.run("你好")
+            return result, synthesize_speech, async_post_message
+
+        result, synthesize_speech, async_post_message = asyncio.run(_run())
+        notification = async_post_message.await_args.args[0]
+
+        self.assertEqual(result, "当前渠道不支持语音回复，已自动回退为文字回复")
+        synthesize_speech.assert_not_called()
+        self.assertEqual(notification.text, "你好")
+        self.assertIsNone(notification.voice_path)
+

 if __name__ == "__main__":
    unittest.main()