livekit-plugins-google 1.1.0__py3-none-any.whl → 1.1.2__py3-none-any.whl

livekit/plugins/google/beta/realtime/realtime_api.py

@@ -68,6 +68,7 @@ class _RealtimeOptions:
     output_audio_transcription: types.AudioTranscriptionConfig | None
     image_encode_options: NotGivenOr[images.EncodeOptions]
     conn_options: APIConnectOptions
+    http_options: NotGivenOr[types.HttpOptions]
     enable_affective_dialog: NotGivenOr[bool] = NOT_GIVEN
     proactivity: NotGivenOr[bool] = NOT_GIVEN
     realtime_input_config: NotGivenOr[types.RealtimeInputConfig] = NOT_GIVEN
@@ -136,6 +137,7 @@ class RealtimeModel(llm.RealtimeModel):
         context_window_compression: NotGivenOr[types.ContextWindowCompressionConfig] = NOT_GIVEN,
         api_version: NotGivenOr[str] = NOT_GIVEN,
         conn_options: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,
+        http_options: NotGivenOr[types.HttpOptions] = NOT_GIVEN,
         _gemini_tools: NotGivenOr[list[_LLMTool]] = NOT_GIVEN,
     ) -> None:
         """
@@ -259,6 +261,7 @@ class RealtimeModel(llm.RealtimeModel):
             api_version=api_version,
             gemini_tools=_gemini_tools,
             conn_options=conn_options,
+            http_options=http_options,
         )
 
         self._sessions = weakref.WeakSet[RealtimeSession]()
@@ -319,7 +322,9 @@ class RealtimeSession(llm.RealtimeSession):
         if not api_version and (self._opts.enable_affective_dialog or self._opts.proactivity):
             api_version = "v1alpha"
 
-        http_options = types.HttpOptions(timeout=int(self._opts.conn_options.timeout * 1000))
+        http_options = self._opts.http_options or types.HttpOptions(
+            timeout=int(self._opts.conn_options.timeout * 1000)
+        )
         if api_version:
             http_options.api_version = api_version
 
@@ -902,6 +907,9 @@ class RealtimeSession(llm.RealtimeSession):
             )
 
         if not gen.text_ch.closed:
+            if self._opts.output_audio_transcription is None:
+                # close the text data of transcription synchronizer
+                gen.text_ch.send_nowait("")
             gen.text_ch.close()
         if not gen.audio_ch.closed:
             gen.audio_ch.close()

livekit/plugins/google/llm.py

@@ -60,6 +60,7 @@ class _LLMOptions:
     frequency_penalty: NotGivenOr[float]
     thinking_config: NotGivenOr[types.ThinkingConfigOrDict]
     gemini_tools: NotGivenOr[list[_LLMTool]]
+    http_options: NotGivenOr[types.HttpOptions]
 
 
 class LLM(llm.LLM):
@@ -80,6 +81,7 @@ class LLM(llm.LLM):
         tool_choice: NotGivenOr[ToolChoice] = NOT_GIVEN,
         thinking_config: NotGivenOr[types.ThinkingConfigOrDict] = NOT_GIVEN,
         gemini_tools: NotGivenOr[list[_LLMTool]] = NOT_GIVEN,
+        http_options: NotGivenOr[types.HttpOptions] = NOT_GIVEN,
     ) -> None:
         """
         Create a new instance of Google GenAI LLM.
@@ -106,6 +108,7 @@ class LLM(llm.LLM):
             tool_choice (ToolChoice, optional): Specifies whether to use tools during response generation. Defaults to "auto".
             thinking_config (ThinkingConfigOrDict, optional): The thinking configuration for response generation. Defaults to None.
             gemini_tools (list[LLMTool], optional): The Gemini-specific tools to use for the session.
+            http_options (HttpOptions, optional): The HTTP options to use for the session.
         """  # noqa: E501
         super().__init__()
         gcp_project = project if is_given(project) else os.environ.get("GOOGLE_CLOUD_PROJECT")
@@ -166,6 +169,7 @@ class LLM(llm.LLM):
             frequency_penalty=frequency_penalty,
             thinking_config=thinking_config,
             gemini_tools=gemini_tools,
+            http_options=http_options,
         )
         self._client = Client(
             api_key=gemini_api_key,
@@ -312,8 +316,9 @@ class LLMStream(llm.LLMStream):
                     if extra_data.system_messages
                     else None
                 ),
-                http_options=types.HttpOptions(
-                    timeout=int(self._conn_options.timeout * 1000),
+                http_options=(
+                    self._llm._opts.http_options
+                    or types.HttpOptions(timeout=int(self._conn_options.timeout * 1000))
                 ),
                 **self._extra_kwargs,
             )

livekit/plugins/google/stt.py

@@ -381,11 +381,14 @@ class SpeechStream(stt.SpeechStream):
         self._reconnect_event.set()
 
     async def _run(self) -> None:
+        audio_pushed = False
+
         # google requires a async generator when calling streaming_recognize
         # this function basically convert the queue into a async generator
         async def input_generator(
             client: SpeechAsyncClient, should_stop: asyncio.Event
         ) -> AsyncGenerator[cloud_speech.StreamingRecognizeRequest, None]:
+            nonlocal audio_pushed
             try:
                 # first request should contain the config
                 yield cloud_speech.StreamingRecognizeRequest(
@@ -402,6 +405,8 @@ class SpeechStream(stt.SpeechStream):
 
                     if isinstance(frame, rtc.AudioFrame):
                         yield cloud_speech.StreamingRecognizeRequest(audio=frame.data.tobytes())
+                        if not audio_pushed:
+                            audio_pushed = True
 
             except Exception:
                 logger.exception("an error occurred while streaming input to google STT")
@@ -470,6 +475,7 @@ class SpeechStream(stt.SpeechStream):
                     has_started = False
 
         while True:
+            audio_pushed = False
             try:
                 async with self._pool.connection(timeout=self._conn_options.timeout) as client:
                     self._streaming_config = cloud_speech.StreamingRecognitionConfig(
@@ -514,13 +520,21 @@ class SpeechStream(stt.SpeechStream):
                             break
                         self._reconnect_event.clear()
                     finally:
-                        await utils.aio.gracefully_cancel(process_stream_task, wait_reconnect_task)
                         should_stop.set()
+                        if not process_stream_task.done() and not wait_reconnect_task.done():
+                            # try to gracefully stop the process_stream_task
+                            try:
+                                await asyncio.wait_for(process_stream_task, timeout=1.0)
+                            except asyncio.TimeoutError:
+                                pass
+
+                        await utils.aio.gracefully_cancel(process_stream_task, wait_reconnect_task)
             except DeadlineExceeded:
                 raise APITimeoutError() from None
             except GoogleAPICallError as e:
                 if e.code == 409:
-                    logger.debug("stream timed out, restarting.")
+                    if audio_pushed:
+                        logger.debug("stream timed out, restarting.")
                 else:
                     raise APIStatusError(
                         f"{e.message} {e.details}", status_code=e.code or -1

livekit/plugins/google/tts.py

@@ -46,6 +46,8 @@ class _TTSOptions:
     effects_profile_id: str
     speaking_rate: float
     tokenizer: tokenize.SentenceTokenizer
+    volume_gain_db: float
+    enable_ssml: bool
 
 
 class TTS(tts.TTS):
@@ -59,12 +61,14 @@ class TTS(tts.TTS):
         pitch: int = 0,
         effects_profile_id: str = "",
         speaking_rate: float = 1.0,
+        volume_gain_db: float = 0.0,
         location: str = "global",
         audio_encoding: texttospeech.AudioEncoding = texttospeech.AudioEncoding.OGG_OPUS,  # type: ignore
         credentials_info: NotGivenOr[dict] = NOT_GIVEN,
         credentials_file: NotGivenOr[str] = NOT_GIVEN,
         tokenizer: NotGivenOr[tokenize.SentenceTokenizer] = NOT_GIVEN,
         use_streaming: bool = True,
+        enable_ssml: bool = False,
     ) -> None:
         """
         Create a new instance of Google TTS.
@@ -82,10 +86,12 @@ class TTS(tts.TTS):
             pitch (float, optional): Speaking pitch, ranging from -20.0 to 20.0 semitones relative to the original pitch. Default is 0.
             effects_profile_id (str): Optional identifier for selecting audio effects profiles to apply to the synthesized speech.
             speaking_rate (float, optional): Speed of speech. Default is 1.0.
+            volume_gain_db (float, optional): Volume gain in decibels. Default is 0.0. In the range [-96.0, 16.0]. Strongly recommended not to exceed +10 (dB).
             credentials_info (dict, optional): Dictionary containing Google Cloud credentials. Default is None.
             credentials_file (str, optional): Path to the Google Cloud credentials JSON file. Default is None.
             tokenizer (tokenize.SentenceTokenizer, optional): Tokenizer for the TTS. Default is a basic sentence tokenizer.
             use_streaming (bool, optional): Whether to use streaming synthesis. Default is True.
+            enable_ssml (bool, optional): Whether to enable SSML support. Default is False.
         """  # noqa: E501
         super().__init__(
             capabilities=tts.TTSCapabilities(streaming=use_streaming),
@@ -93,6 +99,9 @@ class TTS(tts.TTS):
             num_channels=1,
         )
 
+        if enable_ssml and use_streaming:
+            raise ValueError("SSML support is not available for streaming synthesis")
+
         self._client: texttospeech.TextToSpeechAsyncClient | None = None
         self._credentials_info = credentials_info
         self._credentials_file = credentials_file
@@ -118,6 +127,8 @@ class TTS(tts.TTS):
             effects_profile_id=effects_profile_id,
             speaking_rate=speaking_rate,
             tokenizer=tokenizer,
+            volume_gain_db=volume_gain_db,
+            enable_ssml=enable_ssml,
         )
         self._streams = weakref.WeakSet[SynthesizeStream]()
 
@@ -128,6 +139,7 @@ class TTS(tts.TTS):
         gender: NotGivenOr[Gender | str] = NOT_GIVEN,
         voice_name: NotGivenOr[str] = NOT_GIVEN,
         speaking_rate: NotGivenOr[float] = NOT_GIVEN,
+        volume_gain_db: NotGivenOr[float] = NOT_GIVEN,
     ) -> None:
         """
         Update the TTS options.
@@ -137,6 +149,7 @@ class TTS(tts.TTS):
             gender (Gender | str, optional): Voice gender ("male", "female", "neutral").
             voice_name (str, optional): Specific voice name.
             speaking_rate (float, optional): Speed of speech.
+            volume_gain_db (float, optional): Volume gain in decibels.
         """
         params = {}
         if is_given(language):
@@ -151,6 +164,8 @@ class TTS(tts.TTS):
 
         if is_given(speaking_rate):
             self._opts.speaking_rate = speaking_rate
+        if is_given(volume_gain_db):
+            self._opts.volume_gain_db = volume_gain_db
 
     def _ensure_client(self) -> texttospeech.TextToSpeechAsyncClient:
         api_endpoint = "texttospeech.googleapis.com"
@@ -199,10 +214,21 @@ class ChunkedStream(tts.ChunkedStream):
         self._tts: TTS = tts
         self._opts = replace(tts._opts)
 
+    def _build_ssml(self) -> str:
+        ssml = "<speak>"
+        ssml += self._input_text
+        ssml += "</speak>"
+        return ssml
+
     async def _run(self, output_emitter: tts.AudioEmitter) -> None:
         try:
+            input = (
+                texttospeech.SynthesisInput(ssml=self._build_ssml())
+                if self._opts.enable_ssml
+                else texttospeech.SynthesisInput(text=self._input_text)
+            )
             response: SynthesizeSpeechResponse = await self._tts._ensure_client().synthesize_speech(
-                input=texttospeech.SynthesisInput(text=self._input_text),
+                input=input,
                 voice=self._opts.voice,
                 audio_config=texttospeech.AudioConfig(
                     audio_encoding=self._opts.encoding,
@@ -210,6 +236,7 @@ class ChunkedStream(tts.ChunkedStream):
                     pitch=self._opts.pitch,
                     effects_profile_id=self._opts.effects_profile_id,
                     speaking_rate=self._opts.speaking_rate,
+                    volume_gain_db=self._opts.volume_gain_db,
                 ),
                 timeout=self._conn_options.timeout,
             )
@@ -256,7 +283,9 @@ class SynthesizeStream(tts.SynthesizeStream):
         streaming_config = texttospeech.StreamingSynthesizeConfig(
             voice=self._opts.voice,
             streaming_audio_config=texttospeech.StreamingAudioConfig(
-                audio_encoding=encoding, sample_rate_hertz=self._opts.sample_rate
+                audio_encoding=encoding,
+                sample_rate_hertz=self._opts.sample_rate,
+                speaking_rate=self._opts.speaking_rate,
             ),
         )
 

livekit/plugins/google/version.py

@@ -12,4 +12,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "1.1.0"
+__version__ = "1.1.2"

{livekit_plugins_google-1.1.0.dist-info → livekit_plugins_google-1.1.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: livekit-plugins-google
-Version: 1.1.0
+Version: 1.1.2
 Summary: Agent Framework plugin for services from Google Cloud
 Project-URL: Documentation, https://docs.livekit.io
 Project-URL: Website, https://livekit.io/
@@ -20,9 +20,9 @@ Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.9.0
 Requires-Dist: google-auth<3,>=2
 Requires-Dist: google-cloud-speech<3,>=2
-Requires-Dist: google-cloud-texttospeech<3,>=2.24
-Requires-Dist: google-genai>=v1.16.1
-Requires-Dist: livekit-agents>=1.1.0
+Requires-Dist: google-cloud-texttospeech<3,>=2.27
+Requires-Dist: google-genai>=v1.21.1
+Requires-Dist: livekit-agents>=1.1.2
 Description-Content-Type: text/markdown
 
 # Google AI plugin for LiveKit Agents

{livekit_plugins_google-1.1.0.dist-info → livekit_plugins_google-1.1.2.dist-info}/RECORD

@@ -1,17 +1,17 @@
 livekit/plugins/google/__init__.py,sha256=XIyZ-iFnRBpaLtOJgVwojlB-a8GjdDugVFcjBpMEww8,1412
-livekit/plugins/google/llm.py,sha256=MIi-6kk8AZQxcf5y4zB3HwwEQHAJSCIdX79yf9QMAvI,17835
+livekit/plugins/google/llm.py,sha256=Feb2ixNN9YoDt3aPXkQNeVx2c-wkmrf-mv4r3vggY1s,18131
 livekit/plugins/google/log.py,sha256=GI3YWN5YzrafnUccljzPRS_ZALkMNk1i21IRnTl2vNA,69
 livekit/plugins/google/models.py,sha256=hOpfbN_qdQ1ZTpCN9m9dvG2eb6WgQ3KE3WRpIeeM_T0,1569
 livekit/plugins/google/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-livekit/plugins/google/stt.py,sha256=SddM50w6g2rNkjaF5OtrPwEH-qqq36sa-v_6ogKoBYg,24077
+livekit/plugins/google/stt.py,sha256=ssDMH5U1vQOLA44XMlovYWIR4UqVtZSge3YFN-zZ7Iw,24696
 livekit/plugins/google/tools.py,sha256=tD5HVDHO5JfUF029Cx3axHMJec0Gxalkl7s1FDgxLzI,259
-livekit/plugins/google/tts.py,sha256=PzDfEfvQfj-uSHYOUelFnwYK0Wu2-5Mp8PID0b4I5kc,14293
+livekit/plugins/google/tts.py,sha256=YTfce55MWNJyDH4k8U1O2giOcrtccTs8vrkiW9GuBR0,15541
 livekit/plugins/google/utils.py,sha256=-4z6wrjVaZPtFRowkpwaA2acBRfqtzTk4r2xrPDUdCk,8609
-livekit/plugins/google/version.py,sha256=7SjyflIFTjH0djSotKGIRoRykPCqMpVYetIlvHMFuh0,600
+livekit/plugins/google/version.py,sha256=gqaIRup9hxsq6YNsBlKPmS5PL-B8yqSRTd8wRfj8zoQ,600
 livekit/plugins/google/beta/__init__.py,sha256=5PnoG3Ux24bjzMSzmTeSVljE9EINivGcbWUEV6egGnM,216
 livekit/plugins/google/beta/realtime/__init__.py,sha256=_fW2NMN22F-hnQ4xAJ_g5lPbR7CvM_xXzSWlUQY-E-U,188
 livekit/plugins/google/beta/realtime/api_proto.py,sha256=NfE7xr2N3JOu7gVfWbAmDcEhs8vuZgMRu5vpScPJzsg,776
-livekit/plugins/google/beta/realtime/realtime_api.py,sha256=Mt-f7mkwVd7Aq84HPh_AdIOaB4ye8d6TTllcEjKO5TY,45918
-livekit_plugins_google-1.1.0.dist-info/METADATA,sha256=HeQoxgYu0-hOIOawXsvtwHeESXj1U2Oo5GpwEUEx-W8,1907
-livekit_plugins_google-1.1.0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-livekit_plugins_google-1.1.0.dist-info/RECORD,,
+livekit/plugins/google/beta/realtime/realtime_api.py,sha256=tlAsTFsumqOavC9JT2SuQi_3eGYygZ3bbS-nEM7ea8Q,46293
+livekit_plugins_google-1.1.2.dist-info/METADATA,sha256=cAk_E0o73mOJ1wFsuUFzmzW4vZ2B_lbM2O3aZeHoHq4,1907
+livekit_plugins_google-1.1.2.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+livekit_plugins_google-1.1.2.dist-info/RECORD,,