livekit-plugins-google 1.2.6__tar.gz → 1.2.8__tar.gz

{livekit_plugins_google-1.2.6 → livekit_plugins_google-1.2.8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: livekit-plugins-google
-Version: 1.2.6
+Version: 1.2.8
 Summary: Agent Framework plugin for services from Google Cloud
 Project-URL: Documentation, https://docs.livekit.io
 Project-URL: Website, https://livekit.io/
@@ -22,7 +22,7 @@ Requires-Dist: google-auth<3,>=2
 Requires-Dist: google-cloud-speech<3,>=2
 Requires-Dist: google-cloud-texttospeech<3,>=2.27
 Requires-Dist: google-genai>=v1.23.0
-Requires-Dist: livekit-agents>=1.2.6
+Requires-Dist: livekit-agents>=1.2.8
 Description-Content-Type: text/markdown
 
 # Google AI plugin for LiveKit Agents

{livekit_plugins_google-1.2.6 → livekit_plugins_google-1.2.8}/livekit/plugins/google/beta/realtime/realtime_api.py

@@ -708,7 +708,7 @@ class RealtimeSession(llm.RealtimeSession):
                             # reset the flag and still start a new generation in case it has any other content
                             response.server_content.interrupted = False
 
-                        if response.server_content or response.tool_call:
+                        if self._is_new_generation(response):
                             self._start_new_generation()
 
                     if response.session_resumption_update:
@@ -1091,3 +1091,16 @@ class RealtimeSession(llm.RealtimeSession):
                 recoverable=recoverable,
             ),
         )
+
+    def _is_new_generation(self, resp: types.LiveServerMessage) -> bool:
+        if resp.tool_call:
+            return True
+
+        if (sc := resp.server_content) and (
+            sc.model_turn
+            or (sc.output_transcription and sc.output_transcription.text is not None)
+            or (sc.input_transcription and sc.input_transcription.text is not None)
+        ):
+            return True
+
+        return False

{livekit_plugins_google-1.2.6 → livekit_plugins_google-1.2.8}/livekit/plugins/google/llm.py

@@ -157,10 +157,6 @@ class LLM(llm.LLM):
             if _thinking_budget is not None:
                 if not isinstance(_thinking_budget, int):
                     raise ValueError("thinking_budget inside thinking_config must be an integer")
-                if not (0 <= _thinking_budget <= 24576):
-                    raise ValueError(
-                        "thinking_budget inside thinking_config must be between 0 and 24576"
-                    )
 
         self._opts = _LLMOptions(
             model=model,

{livekit_plugins_google-1.2.6 → livekit_plugins_google-1.2.8}/livekit/plugins/google/stt.py

@@ -20,6 +20,7 @@ import time
 import weakref
 from collections.abc import AsyncGenerator, AsyncIterable
 from dataclasses import dataclass
+from datetime import timedelta
 from typing import Callable, Union, cast
 
 from google.api_core.client_options import ClientOptions
@@ -28,6 +29,7 @@ from google.auth import default as gauth_default
 from google.auth.exceptions import DefaultCredentialsError
 from google.cloud.speech_v2 import SpeechAsyncClient
 from google.cloud.speech_v2.types import cloud_speech
+from google.protobuf.duration_pb2 import Duration
 from livekit import rtc
 from livekit.agents import (
     DEFAULT_API_CONNECT_OPTIONS,
@@ -552,6 +554,14 @@ class SpeechStream(stt.SpeechStream):
                 raise APIConnectionError() from e
 
 
+def _duration_to_seconds(duration: Duration | timedelta) -> float:
+    # Proto Plus may auto-convert Duration to timedelta; handle both.
+    # https://proto-plus-python.readthedocs.io/en/latest/marshal.html
+    if isinstance(duration, timedelta):
+        return duration.total_seconds()
+    return duration.seconds + duration.nanos / 1e9
+
+
 def _recognize_response_to_speech_event(
     resp: cloud_speech.RecognizeResponse,
 ) -> stt.SpeechEvent:
@@ -561,24 +571,31 @@ def _recognize_response_to_speech_event(
         text += result.alternatives[0].transcript
         confidence += result.alternatives[0].confidence
 
-    # not sure why start_offset and end_offset returns a timedelta
-    start_offset = resp.results[0].alternatives[0].words[0].start_offset
-    end_offset = resp.results[-1].alternatives[0].words[-1].end_offset
+    alternatives = []
 
-    confidence /= len(resp.results)
-    lg = resp.results[0].language_code
-    return stt.SpeechEvent(
-        type=stt.SpeechEventType.FINAL_TRANSCRIPT,
-        alternatives=[
+    # Google STT may return empty results when spoken_lang != stt_lang
+    if resp.results:
+        try:
+            start_time = _duration_to_seconds(resp.results[0].alternatives[0].words[0].start_offset)
+            end_time = _duration_to_seconds(resp.results[-1].alternatives[0].words[-1].end_offset)
+        except IndexError:
+            # When enable_word_time_offsets=False, there are no "words" to access
+            start_time = end_time = 0
+
+        confidence /= len(resp.results)
+        lg = resp.results[0].language_code
+
+        alternatives = [
             stt.SpeechData(
                 language=lg,
-                start_time=start_offset.total_seconds(),  # type: ignore
-                end_time=end_offset.total_seconds(),  # type: ignore
+                start_time=start_time,
+                end_time=end_time,
                 confidence=confidence,
                 text=text,
             )
-        ],
-    )
+        ]
+
+    return stt.SpeechEvent(type=stt.SpeechEventType.FINAL_TRANSCRIPT, alternatives=alternatives)
 
 
 def _streaming_recognize_response_to_speech_data(

{livekit_plugins_google-1.2.6 → livekit_plugins_google-1.2.8}/livekit/plugins/google/tts.py

@@ -124,7 +124,7 @@ class TTS(tts.TTS):
         )
         if is_given(voice_cloning_key):
             voice_params.voice_clone = texttospeech.VoiceCloneParams(
-                voice_clone_key=voice_cloning_key,
+                voice_cloning_key=voice_cloning_key,
             )
         else:
             voice_params.name = voice_name if is_given(voice_name) else DEFAULT_VOICE_NAME

{livekit_plugins_google-1.2.6 → livekit_plugins_google-1.2.8}/livekit/plugins/google/version.py

@@ -12,4 +12,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "1.2.6"
+__version__ = "1.2.8"

{livekit_plugins_google-1.2.6 → livekit_plugins_google-1.2.8}/pyproject.toml

@@ -27,7 +27,7 @@ dependencies = [
     "google-cloud-speech >= 2, < 3",
     "google-cloud-texttospeech >= 2.27, < 3",
     "google-genai >= v1.23.0",
-    "livekit-agents>=1.2.6",
+    "livekit-agents>=1.2.8",
 ]
 
 [project.urls]