livekit-plugins-volcengine 1.2.3.post0__tar.gz → 1.2.6__tar.gz

{livekit_plugins_volcengine-1.2.3.post0 → livekit_plugins_volcengine-1.2.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: livekit-plugins-volcengine
-Version: 1.2.3.post0
+Version: 1.2.6
 Summary: LiveKit Agent Plugins for Volcengine
 Author-email: wangmengdi <790990241@qq.com>
 Keywords: audio,livekit,realtime,video,webrtc
@@ -14,7 +14,7 @@ Classifier: Topic :: Multimedia :: Sound/Audio
 Classifier: Topic :: Multimedia :: Video
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.9
-Requires-Dist: livekit-agents==1.2.2
+Requires-Dist: livekit-agents>=1.2.6
 Requires-Dist: numpy
 Requires-Dist: openai>=1.75.0
 Requires-Dist: osc-data>=0.2.2

{livekit_plugins_volcengine-1.2.3.post0 → livekit_plugins_volcengine-1.2.6}/livekit/plugins/volcengine/realtime.py

@@ -169,6 +169,7 @@ class _RealtimeOptions:
     system_role: str
     max_session_duration: float | None
     conn_options: APIConnectOptions
+    modalities: list[Literal["text", "audio"]]
     opening: str = "你好啊，今天过得怎么样？"
     speaking_style: str = "你的说话风格简洁明了，语速适中，语调自然。"
     speaker: str = "zh_female_vv_jupiter_bigtts"
@@ -216,8 +217,8 @@ class _MessageGeneration:
     message_id: str
     text_ch: utils.aio.Chan[str]
     audio_ch: utils.aio.Chan[rtc.AudioFrame]
-    modalities: asyncio.Future[list[Literal["text", "audio"]]]
     audio_transcript: str = ""
+    modalities: asyncio.Future[list[Literal["text", "audio"]]] | None = None
 
 
 @dataclass
@@ -245,17 +246,19 @@ class RealtimeModel(llm.RealtimeModel):
         access_token: str | None = None,
         system_role: str | None = None,
         audio_output: bool = True,
+        modalities: NotGivenOr[list[Literal["text", "audio"]]] = NOT_GIVEN,
         http_session: aiohttp.ClientSession | None = None,
         max_session_duration: NotGivenOr[float | None] = NOT_GIVEN,
         conn_options: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,
     ) -> None:
+        modalities = modalities if utils.is_given(modalities) else ["text", "audio"]
         super().__init__(
             capabilities=llm.RealtimeCapabilities(
                 message_truncation=True,
                 turn_detection=True,
                 user_transcription=True,
                 auto_tool_reply_generation=False,
-                audio_output=audio_output,
+                audio_output=("audio" in modalities),
             )
         )
 
@@ -277,10 +280,10 @@ class RealtimeModel(llm.RealtimeModel):
             speaker=speaker,
             opening=opening,
             speaking_style=speaking_style,
+            modalities=modalities,
         )
         self._http_session = http_session
         self._sessions = weakref.WeakSet[RealtimeSession]()
-        print(self.capabilities.audio_output)
 
     def update_options(
         self,
@@ -438,15 +441,18 @@ class RealtimeSession(
             )
             self.emit("generation_created", generation_ev)
             item_id = utils.shortuuid()
+            modalities_fut: asyncio.Future[list[Literal["text", "audio"]]] = asyncio.Future()
             self._current_item = _MessageGeneration(
                 message_id=item_id,
                 text_ch=utils.aio.Chan(),
                 audio_ch=utils.aio.Chan(),
-                modalities=asyncio.Future(),
+                modalities=modalities_fut,
             )
             if not self._realtime_model.capabilities.audio_output:
                 self._current_item.audio_ch.close()
-                self._current_item.modalities.set_result(["text"])
+                self._current_item.modalities.set_result(["text"])  # type: ignore[union-attr]
+            else:
+                self._current_item.modalities.set_result(["audio", "text"])  # type: ignore[union-attr]
 
             self._current_generation.message_ch.send_nowait(
                 llm.MessageGeneration(
@@ -505,7 +511,6 @@ class RealtimeSession(
                         is_final = not response["results"][0]["is_interim"]
                         if is_final:
                             item_id = utils.shortuuid()
-                            logger.info("transcription completed")
                             self.emit(
                                 "input_audio_transcription_completed",
                                 llm.InputTranscriptionCompleted(
@@ -531,15 +536,18 @@ class RealtimeSession(
 
                                 self.emit("generation_created", generation_ev)
                                 item_id = utils.shortuuid()
+                                modalities_fut: asyncio.Future[list[Literal["text", "audio"]]] = asyncio.Future()
                                 self._current_item = _MessageGeneration(
                                     message_id=item_id,
                                     text_ch=utils.aio.Chan(),
                                     audio_ch=utils.aio.Chan(),
-                                    modalities=asyncio.Future(),
+                                    modalities=modalities_fut,
                                 )
                                 if not self._realtime_model.capabilities.audio_output:
                                     self._current_item.audio_ch.close()
-                                    self._current_item.modalities.set_result(["text"])
+                                    self._current_item.modalities.set_result(["text"])  # type: ignore[union-attr]
+                                else:
+                                    self._current_item.modalities.set_result(["audio", "text"])  # type: ignore[union-attr]
                                 self._current_generation.message_ch.send_nowait(
                                     llm.MessageGeneration(
                                         message_id=item_id,
@@ -561,7 +569,6 @@ class RealtimeSession(
                         logger.info("tts start")
 
                     elif event == 352:  # TTSResponse
-                        logger.info("tts response")
                         if self._first_tts_response:
                             logger.info("llm first sentence")
                             logger.info("tts first response")
@@ -766,13 +773,15 @@ class RealtimeSession(
         self,
         *,
         message_id: str,
+        modalities: list[Literal["text", "audio"]],
         audio_end_ms: int,
         audio_transcript: NotGivenOr[str] = NOT_GIVEN,
     ) -> None:
-        if self._realtime_model.capabilities.audio_output:
+        if "audio" in modalities:
+            # 当前 volcengine 实时接口未暴露远端音频截断事件；占位以对齐接口
             pass
         elif utils.is_given(audio_transcript):
-            # sync the forwarded text to the remote chat ctx
+            # 同步转写文本到远端会话上下文
             chat_ctx = self.chat_ctx.copy()
             if (idx := chat_ctx.index_by_id(message_id)) is not None:
                 new_item = copy.copy(chat_ctx.items[idx])

livekit_plugins_volcengine-1.2.6/livekit/plugins/volcengine/version.py

@@ -0,0 +1 @@
+__version__ = "1.2.6"

{livekit_plugins_volcengine-1.2.3.post0 → livekit_plugins_volcengine-1.2.6}/pyproject.toml

@@ -13,7 +13,7 @@ dependencies = [
     "osc-data>=0.2.2",
     "pydantic",
     "numpy",
-    "livekit-agents==1.2.2",
+    "livekit-agents>=1.2.6",
 ]
 
 classifiers = [

livekit_plugins_volcengine-1.2.3.post0/livekit/plugins/volcengine/version.py

@@ -1 +0,0 @@
-__version__ = "1.2.3.post0"