livekit-plugins-google 0.11.1__py3-none-any.whl → 1.0.0.dev5__py3-none-any.whl

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

@@ -1,8 +1,4 @@
-from .api_proto import (
-    ClientEvents,
-    LiveAPIModels,
-    Voice,
-)
+from .api_proto import ClientEvents, LiveAPIModels, Voice
 from .realtime_api import RealtimeModel
 
 __all__ = [

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

@@ -1,12 +1,13 @@
 from __future__ import annotations
 
-from typing import Literal, Sequence, Union
+from collections.abc import Sequence
+from typing import Literal, Union
 
 from google.genai import types
 
 from ..._utils import _build_gemini_ctx, _build_tools
 
-LiveAPIModels = Literal["gemini-2.0-flash-exp"]
+LiveAPIModels = Literal["gemini-2.0-flash-001",]
 
 Voice = Literal["Puck", "Charon", "Kore", "Fenrir", "Aoede"]
 

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

@@ -3,21 +3,17 @@ from __future__ import annotations
 import asyncio
 import json
 import os
+from collections.abc import AsyncIterable
 from dataclasses import dataclass
-from typing import AsyncIterable, Literal
-
-from livekit import rtc
-from livekit.agents import llm, utils
-from livekit.agents.llm.function_context import _create_ai_function_info
-from livekit.agents.utils import images
+from typing import Literal
 
 from google import genai
+from google.genai._api_client import HttpOptions
 from google.genai.types import (
     Blob,
     Content,
     FunctionResponse,
     GenerationConfig,
-    HttpOptions,
     LiveClientContent,
     LiveClientRealtimeInput,
     LiveClientToolResponse,
@@ -29,15 +25,13 @@ from google.genai.types import (
     Tool,
     VoiceConfig,
 )
+from livekit import rtc
+from livekit.agents import llm, utils
+from livekit.agents.llm.function_context import _create_ai_function_info
+from livekit.agents.utils import images
 
 from ...log import logger
-from .api_proto import (
-    ClientEvents,
-    LiveAPIModels,
-    Voice,
-    _build_gemini_ctx,
-    _build_tools,
-)
+from .api_proto import ClientEvents, LiveAPIModels, Voice, _build_gemini_ctx, _build_tools
 from .transcriber import ModelTranscriber, TranscriberSession, TranscriptionContent
 
 EventTypes = Literal[
@@ -83,7 +77,6 @@ class Capabilities:
 class ModelOptions:
     model: LiveAPIModels | str
     api_key: str | None
-    api_version: str
     voice: Voice | str
     response_modalities: list[Modality] | None
     vertexai: bool
@@ -108,9 +101,8 @@ class RealtimeModel:
         instructions: str | None = None,
         model: LiveAPIModels | str = "gemini-2.0-flash-exp",
         api_key: str | None = None,
-        api_version: str = "v1alpha",
         voice: Voice | str = "Puck",
-        modalities: list[Modality] = [Modality.AUDIO],
+        modalities: list[Modality] = None,
         enable_user_audio_transcription: bool = True,
         enable_agent_audio_transcription: bool = True,
         vertexai: bool = False,
@@ -138,7 +130,6 @@ class RealtimeModel:
         Args:
             instructions (str, optional): Initial system instructions for the model. Defaults to "".
             api_key (str or None, optional): Google Gemini API key. If None, will attempt to read from the environment variable GOOGLE_API_KEY.
-            api_version (str, optional): The version of the API to use. Defaults to "v1alpha".
             modalities (list[Modality], optional): Modalities to use, such as ["TEXT", "AUDIO"]. Defaults to ["AUDIO"].
             model (str or None, optional): The name of the model to use. Defaults to "gemini-2.0-flash-exp".
             voice (api_proto.Voice, optional): Voice setting for audio outputs. Defaults to "Puck".
@@ -158,6 +149,8 @@ class RealtimeModel:
         Raises:
             ValueError: If the API key is not provided and cannot be found in environment variables.
         """
+        if modalities is None:
+            modalities = ["AUDIO"]
         super().__init__()
         self._capabilities = Capabilities(
             supports_truncate=False,
@@ -183,14 +176,11 @@ class RealtimeModel:
                     "API key is required for Google API either via api_key or GOOGLE_API_KEY environment variable"
                 )
 
-        instructions_content = (
-            Content(parts=[Part(text=instructions)]) if instructions else None
-        )
+        instructions_content = Content(parts=[Part(text=instructions)]) if instructions else None
 
         self._rt_sessions: list[GeminiRealtimeSession] = []
         self._opts = ModelOptions(
             model=model,
-            api_version=api_version,
             api_key=self._api_key,
             voice=voice,
             enable_user_audio_transcription=enable_user_audio_transcription,
@@ -263,8 +253,6 @@ class GeminiRealtimeSession(utils.EventEmitter[EventTypes]):
         self._fnc_ctx = fnc_ctx
         self._fnc_tasks = utils.aio.TaskSet()
         self._is_interrupted = False
-        self._playout_complete = asyncio.Event()
-        self._playout_complete.set()
 
         tools = []
         if self._fnc_ctx is not None:
@@ -285,32 +273,24 @@ class GeminiRealtimeSession(utils.EventEmitter[EventTypes]):
             system_instruction=self._opts.instructions,
             speech_config=SpeechConfig(
                 voice_config=VoiceConfig(
-                    prebuilt_voice_config=PrebuiltVoiceConfig(
-                        voice_name=self._opts.voice
-                    )
+                    prebuilt_voice_config=PrebuiltVoiceConfig(voice_name=self._opts.voice)
                 )
             ),
             tools=tools,
         )
         self._client = genai.Client(
-            http_options=HttpOptions(api_version=self._opts.api_version),
+            http_options=HttpOptions(api_version="v1alpha"),
             api_key=self._opts.api_key,
             vertexai=self._opts.vertexai,
             project=self._opts.project,
             location=self._opts.location,
         )
-        self._main_atask = asyncio.create_task(
-            self._main_task(), name="gemini-realtime-session"
-        )
+        self._main_atask = asyncio.create_task(self._main_task(), name="gemini-realtime-session")
         if self._opts.enable_user_audio_transcription:
-            self._transcriber = TranscriberSession(
-                client=self._client, model=self._opts.model
-            )
+            self._transcriber = TranscriberSession(client=self._client, model=self._opts.model)
             self._transcriber.on("input_speech_done", self._on_input_speech_done)
         if self._opts.enable_agent_audio_transcription:
-            self._agent_transcriber = ModelTranscriber(
-                client=self._client, model=self._opts.model
-            )
+            self._agent_transcriber = ModelTranscriber(client=self._client, model=self._opts.model)
             self._agent_transcriber.on("input_speech_done", self._on_agent_speech_done)
         # init dummy task
         self._init_sync_task = asyncio.create_task(asyncio.sleep(0))
@@ -324,10 +304,6 @@ class GeminiRealtimeSession(utils.EventEmitter[EventTypes]):
         self._send_ch.close()
         await self._main_atask
 
-    @property
-    def playout_complete(self) -> asyncio.Event | None:
-        return self._playout_complete
-
     @property
     def fnc_ctx(self) -> llm.FunctionContext | None:
         return self._fnc_ctx
@@ -345,9 +321,7 @@ class GeminiRealtimeSession(utils.EventEmitter[EventTypes]):
     DEFAULT_ENCODE_OPTIONS = images.EncodeOptions(
         format="JPEG",
         quality=75,
-        resize_options=images.ResizeOptions(
-            width=1024, height=1024, strategy="scale_aspect_fit"
-        ),
+        resize_options=images.ResizeOptions(width=1024, height=1024, strategy="scale_aspect_fit"),
     )
 
     def push_video(
@@ -397,9 +371,7 @@ class GeminiRealtimeSession(utils.EventEmitter[EventTypes]):
 
     def create_response(
         self,
-        on_duplicate: Literal[
-            "cancel_existing", "cancel_new", "keep_both"
-        ] = "keep_both",
+        on_duplicate: Literal["cancel_existing", "cancel_new", "keep_both"] = "keep_both",
     ) -> None:
         turns, _ = _build_gemini_ctx(self._chat_ctx, id(self))
         ctx = [self._opts.instructions] + turns if self._opts.instructions else turns
@@ -485,8 +457,7 @@ class GeminiRealtimeSession(utils.EventEmitter[EventTypes]):
                                         data=part.inline_data.data,
                                         sample_rate=24000,
                                         num_channels=1,
-                                        samples_per_channel=len(part.inline_data.data)
-                                        // 2,
+                                        samples_per_channel=len(part.inline_data.data) // 2,
                                     )
                                     if self._opts.enable_agent_audio_transcription:
                                         content.audio.append(frame)
@@ -529,12 +500,12 @@ class GeminiRealtimeSession(utils.EventEmitter[EventTypes]):
                         logger.warning(
                             "function call cancelled",
                             extra={
-                                "function_call_ids": response.tool_call_cancellation.ids,
+                                "function_call_ids": response.tool_call_cancellation.function_call_ids,
                             },
                         )
                         self.emit(
                             "function_calls_cancelled",
-                            response.tool_call_cancellation.ids,
+                            response.tool_call_cancellation.function_call_ids,
                         )
 
         async with self._client.aio.live.connect(

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

@@ -6,12 +6,12 @@ from dataclasses import dataclass
 from typing import Literal
 
 import websockets
-from livekit import rtc
-from livekit.agents import APIConnectionError, APIStatusError, utils
 
 from google import genai
 from google.genai import types
 from google.genai.errors import APIError, ClientError, ServerError
+from livekit import rtc
+from livekit.agents import APIConnectionError, APIStatusError, utils
 
 from ...log import logger
 from .api_proto import ClientEvents, LiveAPIModels
@@ -51,11 +51,9 @@ class TranscriberSession(utils.EventEmitter[EventTypes]):
         self._needed_sr = 16000
         self._closed = False
 
-        system_instructions = types.Content(
-            parts=[types.Part(text=SYSTEM_INSTRUCTIONS)]
-        )
+        system_instructions = types.Content(parts=[types.Part(text=SYSTEM_INSTRUCTIONS)])
         self._config = types.LiveConnectConfig(
-            response_modalities=[types.Modality.TEXT],
+            response_modalities=["TEXT"],
             system_instruction=system_instructions,
             generation_config=types.GenerationConfig(temperature=0.0),
         )
@@ -81,17 +79,13 @@ class TranscriberSession(utils.EventEmitter[EventTypes]):
             for f in self._resampler.push(frame):
                 self._queue_msg(
                     types.LiveClientRealtimeInput(
-                        media_chunks=[
-                            types.Blob(data=f.data.tobytes(), mime_type="audio/pcm")
-                        ]
+                        media_chunks=[types.Blob(data=f.data.tobytes(), mime_type="audio/pcm")]
                     )
                 )
         else:
             self._queue_msg(
                 types.LiveClientRealtimeInput(
-                    media_chunks=[
-                        types.Blob(data=frame.data.tobytes(), mime_type="audio/pcm")
-                    ]
+                    media_chunks=[types.Blob(data=frame.data.tobytes(), mime_type="audio/pcm")]
                 )
             )
 
@@ -157,17 +151,11 @@ class TranscriberSession(utils.EventEmitter[EventTypes]):
                 logger.exception(f"Uncaught error in transcriber _recv_task: {e}")
                 self._closed = True
 
-        async with self._client.aio.live.connect(
-            model=self._model, config=self._config
-        ) as session:
+        async with self._client.aio.live.connect(model=self._model, config=self._config) as session:
             self._session = session
             tasks = [
-                asyncio.create_task(
-                    _send_task(), name="gemini-realtime-transcriber-send"
-                ),
-                asyncio.create_task(
-                    _recv_task(), name="gemini-realtime-transcriber-recv"
-                ),
+                asyncio.create_task(_send_task(), name="gemini-realtime-transcriber-send"),
+                asyncio.create_task(_recv_task(), name="gemini-realtime-transcriber-recv"),
             ]
 
             try:
@@ -187,9 +175,7 @@ class ModelTranscriber(utils.EventEmitter[EventTypes]):
         self._client = client
         self._model = model
         self._needed_sr = 16000
-        self._system_instructions = types.Content(
-            parts=[types.Part(text=SYSTEM_INSTRUCTIONS)]
-        )
+        self._system_instructions = types.Content(parts=[types.Part(text=SYSTEM_INSTRUCTIONS)])
         self._config = types.GenerateContentConfig(
             temperature=0.0,
             system_instruction=self._system_instructions,
@@ -198,9 +184,7 @@ class ModelTranscriber(utils.EventEmitter[EventTypes]):
         self._resampler: rtc.AudioResampler | None = None
         self._buffer: rtc.AudioFrame | None = None
         self._audio_ch = utils.aio.Chan[rtc.AudioFrame]()
-        self._main_atask = asyncio.create_task(
-            self._main_task(), name="gemini-model-transcriber"
-        )
+        self._main_atask = asyncio.create_task(self._main_task(), name="gemini-model-transcriber")
 
     async def aclose(self) -> None:
         if self._audio_ch.closed: