[yui0303]

AS IS

• BackgroundAudioPlayer can pick the same thinking sound from a list twice
  in a row, so repeated thinking states play an identical filler.

TO BE

• _select_sound_from_list / _normalize_sound_source / play gain an exclude
  arg; _agent_state_changed normalizes once excluding the last played sound,
  plays it, and records it in _last_thinking_sound.

Reimplemented on upstream 1.6.3 (not cherry-picked) from CeranaStudio/agents
f700a5d (orig author Lanznx); adapted to 1.6.3 _normalize_sound_source
returning AudioConfig instead of a (source, volume) tuple.

Co-Authored-By: Claude Opus 4.8 (1M context) noreply@anthropic.com

[github-actions]

STT Test Results

Status: ✗ Some tests failed

Metric	Count
✓ Passed	0
✗ Failed	5
× Errors	0
→ Skipped	40
▣ Total	45
⏱ Duration	129.4s

Failed Tests

• tests.test_stt::test_recognize[livekit.plugins.openai]

  stt_factory = <function parameter_factory.<locals>.<lambda> at 0x7fa9df42fce0>
  request = <FixtureRequest for <Coroutine test_recognize[livekit.plugins.openai]>>
  
      @pytest.mark.usefixtures("job_process")
      @pytest.mark.parametrize("stt_factory", STTs)
      async def test_recognize(stt_factory: Callable[[], stt.STT], request):
          plugin_id = request.node.callspec.id.split("-")[0]  # e.g., "livekit.plugins.deepgram"
          sample_rate = SAMPLE_RATE
          frames, transcript, duration = await make_test_speech(sample_rate=sample_rate)
    
          # TODO: differentiate missing key vs other errors
          try:
  >           stt_instance = stt_factory()
                             ^^^^^^^^^^^^^
  
  tests/test_stt.py:153: 
  _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
  
  >   return pytest.param(lambda: plugin.STT(), id=plugin.__name__)
                                  ^^^^^^^^^^^^
  
  tests/test_stt.py:48: 
  _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
  
  self = <livekit.plugins.openai.stt.STT object at 0x7fa9df415070>
  language = 'en', detect_language = False, model = 'gpt-4o-mini-transcribe'
  prompt = NOT_GIVEN, keywords = NOT_GIVEN
  turn_detection = {'type': 'server_vad', 'threshold': 0.5, 'prefix_padding_ms': 600, 'silence_duration_ms': 350}
  noise_reduction_type = NOT_GIVEN, temperature = NOT_GIVEN, base_url = NOT_GIVEN
  api_key = NOT_GIVEN, client = None, use_realtime = False, vad = NOT_GIVEN
  
      def __init__(
          self,
          *,
          language: str | list[str] = "en",
          detect_language: bool = False,
          model: STTModels | str = "gpt-4o-mini-transcribe",
          prompt: NotGivenOr[str] = NOT_GIVEN,
          keywords: NotGivenOr[list[str]] = NOT_GIVEN,
          turn_detection: NotGivenOr[SessionTurnDetection] = NOT_GIVEN,
          noise_reduction_type: NotGivenOr[str] = NOT_GIVEN,
          temperature: NotGivenOr[float] = NOT_GIVEN,
          base_url: NotGivenOr[str] = NOT_GIVEN,
          api_ke
  

• tests.test_stt::test_stream[livekit.plugins.fireworksai]

  self = <livekit.plugins.fireworksai.stt.SpeechStream object at 0x7fe894141250>
  
      async def __anext__(self) -> SpeechEvent:
          try:
  >           val = await self._event_aiter.__anext__()
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  E           StopAsyncIteration
  
  livekit-agents/livekit/agents/stt/stt.py:613: StopAsyncIteration
  
  During handling of the above exception, another exception occurred:
  
  stt_factory = <function parameter_factory.<locals>.<lambda> at 0x7fe886d64900>
  request = <FixtureRequest for <Coroutine test_stream[livekit.plugins.fireworksai]>>
  
      @pytest.mark.usefixtures("job_process")
      @pytest.mark.parametrize("stt_factory", STTs + STREAM_ONLY_STTs)
      async def test_stream(stt_factory: Callable[[], STT], request):
          sample_rate = SAMPLE_RATE
          plugin_id = request.node.callspec.id.split("-")[0]
          frames, transcript, _ = await make_test_speech(
              chunk_duration_ms=STREAM_CHUNK_DURATION_MS, sample_rate=sample_rate
          )
    
          # TODO: differentiate missing key vs other errors
          try:
              stt_instance: STT = stt_factory()
          except ValueError as e:
              pytest.skip(f"{plugin_id}: {e}")
    
          async with stt_instance as stt:
              label = f"{stt.model}@{stt.provider}"
              if not stt.capabilities.streaming:
                  pytest.skip(f"{label} does not support streaming")
    
              for attempt in range(MAX_RETRIES):
                  try:
                      state = {"closing": False}
    
                      async def _stream_input(
                          frames: list[rtc.AudioFrame], stream: RecognizeStream, state: dict = state
                      ):
                          for frame in frames:
                              stream.push_frame(frame)
                              await asyncio.sleep(frame.duration)
    
                          silence = rtc.AudioFrame.create(
                              sample_rate=sample_rate,
  

• tests.test_stt::test_stream[livekit.plugins.openai]

  stt_factory = <function parameter_factory.<locals>.<lambda> at 0x7fa9df42fce0>
  request = <FixtureRequest for <Coroutine test_stream[livekit.plugins.openai]>>
  
      @pytest.mark.usefixtures("job_process")
      @pytest.mark.parametrize("stt_factory", STTs + STREAM_ONLY_STTs)
      async def test_stream(stt_factory: Callable[[], STT], request):
          sample_rate = SAMPLE_RATE
          plugin_id = request.node.callspec.id.split("-")[0]
          frames, transcript, _ = await make_test_speech(
              chunk_duration_ms=STREAM_CHUNK_DURATION_MS, sample_rate=sample_rate
          )
    
          # TODO: differentiate missing key vs other errors
          try:
  >           stt_instance: STT = stt_factory()
                                  ^^^^^^^^^^^^^
  
  tests/test_stt.py:208: 
  _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
  
  >   return pytest.param(lambda: plugin.STT(), id=plugin.__name__)
                                  ^^^^^^^^^^^^
  
  tests/test_stt.py:48: 
  _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
  
  self = <livekit.plugins.openai.stt.STT object at 0x7fa9ec2b4830>
  language = 'en', detect_language = False, model = 'gpt-4o-mini-transcribe'
  prompt = NOT_GIVEN, keywords = NOT_GIVEN
  turn_detection = {'type': 'server_vad', 'threshold': 0.5, 'prefix_padding_ms': 600, 'silence_duration_ms': 350}
  noise_reduction_type = NOT_GIVEN, temperature = NOT_GIVEN, base_url = NOT_GIVEN
  api_key = NOT_GIVEN, client = None, use_realtime = False, vad = NOT_GIVEN
  
      def __init__(
          self,
          *,
          language: str | list[str] = "en",
          detect_language: bool = False,
          model: STTModels | str = "gpt-4o-mini-transcribe",
          prompt: NotGivenOr[str] = NOT_GIVEN,
          keywords: NotGivenOr[list[str]] = NOT_GIVEN,
          turn_detection: NotGivenOr[SessionTurnDetection] = NOT_GIVEN,
          noise_reduction_type: NotGivenOr[str] = NOT_GIVEN,
          temperature: NotGivenOr[float] = NOT_GIVEN,
          base_url: N
  

• tests.test_stt::test_stream[livekit.plugins.aws]

  stt_factory = <function parameter_factory.<locals>.<lambda> at 0x7f0fe6c387c0>
  request = <FixtureRequest for <Coroutine test_stream[livekit.plugins.aws]>>
  
      @pytest.mark.usefixtures("job_process")
      @pytest.mark.parametrize("stt_factory", STTs + STREAM_ONLY_STTs)
      async def test_stream(stt_factory: Callable[[], STT], request):
          sample_rate = SAMPLE_RATE
          plugin_id = request.node.callspec.id.split("-")[0]
          frames, transcript, _ = await make_test_speech(
              chunk_duration_ms=STREAM_CHUNK_DURATION_MS, sample_rate=sample_rate
          )
    
          # TODO: differentiate missing key vs other errors
          try:
              stt_instance: STT = stt_factory()
          except ValueError as e:
              pytest.skip(f"{plugin_id}: {e}")
    
          async with stt_instance as stt:
              label = f"{stt.model}@{stt.provider}"
              if not stt.capabilities.streaming:
                  pytest.skip(f"{label} does not support streaming")
    
              for attempt in range(MAX_RETRIES):
                  try:
                      state = {"closing": False}
    
                      async def _stream_input(
                          frames: list[rtc.AudioFrame], stream: RecognizeStream, state: dict = state
                      ):
                          for frame in frames:
                              stream.push_frame(frame)
                              await asyncio.sleep(frame.duration)
    
                          silence = rtc.AudioFrame.create(
                              sample_rate=sample_rate,
                              num_channels=1,
                              samples_per_channel=sample_rate * STREAM_CHUNK_DURATION_MS // 1000,
                          )
                          for _ in range(TRAILING_SILENCE_DURATION_MS // STREAM_CHUNK_DURATION_MS):
                              stream.push_frame(silence)
                              await asyncio.sleep(silence.duration)
    
                          stream.end_input()
  

• tests.test_stt::test_stream[livekit.plugins.openai.realtime]

  stt_factory = <function <lambda> at 0x7fa9df4344a0>
  request = <FixtureRequest for <Coroutine test_stream[livekit.plugins.openai.realtime]>>
  
      @pytest.mark.usefixtures("job_process")
      @pytest.mark.parametrize("stt_factory", STTs + STREAM_ONLY_STTs)
      async def test_stream(stt_factory: Callable[[], STT], request):
          sample_rate = SAMPLE_RATE
          plugin_id = request.node.callspec.id.split("-")[0]
          frames, transcript, _ = await make_test_speech(
              chunk_duration_ms=STREAM_CHUNK_DURATION_MS, sample_rate=sample_rate
          )
    
          # TODO: differentiate missing key vs other errors
          try:
  >           stt_instance: STT = stt_factory()
                                  ^^^^^^^^^^^^^
  
  tests/test_stt.py:208: 
  _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
  
  >       pytest.param(lambda: openai.STT(use_realtime=True), id="livekit.plugins.openai.realtime"),
                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
          pytest.param(
              lambda: inference.STT(model="deepgram/nova-3"), id="livekit.agents.inference.deepgram"
          ),
          pytest.param(
              lambda: inference.STT(model="cartesia/ink-whisper"),
              id="livekit.agents.inference.cartesia",
          ),
          pytest.param(
              lambda: inference.STT(
                  model="assemblyai/universal-streaming", extra_kwargs={"format_turns": True}
              ),
              id="livekit.agents.inference.assemblyai",
          ),
          pytest.param(lambda: inference.STT(model="xai/stt-1"), id="livekit.agents.inference.xai"),
      ]
  
  tests/test_stt.py:88: 
  _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
  
  self = <livekit.plugins.openai.stt.STT object at 0x7fa9dc644200>
  language = 'en', detect_language = False, model = 'gpt-4o-mini-transcribe'
  prompt = NOT_GIVEN, keywords = NOT_GIVEN
  turn_detection = {'type': 'server_vad', 'threshold': 0.5, 'prefix_padding_ms': 600, 'silence_duration_m
  

Skipped Tests

Test	Reason
tests.test_stt::test_recognize[livekit.plugins.deepgram]	livekit.plugins.deepgram: Deepgram API key is required, either as argument or set DEEPGRAM_API_KEY environment variable
tests.test_stt::test_recognize[livekit.plugins.elevenlabs]	livekit.plugins.elevenlabs: ElevenLabs API key is required, either as argument or set ELEVEN_API_KEY environmental variable
tests.test_stt::test_recognize[livekit.plugins.fal]	livekit.plugins.fal: fal AI API key is required. It should be set with env FAL_KEY
tests.test_stt::test_recognize[livekit.plugins.mistralai]	livekit.plugins.mistralai: Mistral AI API key is required. Set MISTRAL_API_KEY or pass api_key
tests.test_stt::test_recognize[livekit.plugins.assemblyai]	livekit.plugins.assemblyai: AssemblyAI API key is required. Pass one in via the api_key parameter, or set it as the ASSEMBLYAI_API_KEY environment variable
tests.test_stt::test_recognize[livekit.plugins.fireworksai]	unknown@FireworksAI does not support batch recognition
tests.test_stt::test_recognize[livekit.plugins.nvidia]	livekit.plugins.nvidia: NVIDIA_API_KEY is not set while using SSL. Either pass api_key parameter, set NVIDIA_API_KEY environment variable or disable SSL and use a locally hosted Riva NIM service.
tests.test_stt::test_recognize[livekit.plugins.speechmatics]	livekit.plugins.speechmatics: Missing Speechmatics API key
tests.test_stt::test_recognize[livekit.plugins.gladia]	livekit.plugins.gladia: Gladia API key is required. Set GLADIA_API_KEY or pass api_key
tests.test_stt::test_recognize[livekit.plugins.deepgram.STTv2]	livekit.plugins.deepgram.STTv2: Deepgram API key is required
tests.test_stt::test_recognize[livekit.plugins.azure]	livekit.plugins.azure: AZURE_SPEECH_HOST or AZURE_SPEECH_KEY and AZURE_SPEECH_REGION or speech_auth_token and AZURE_SPEECH_REGION or AZURE_SPEECH_KEY and speech_endpoint must be set
tests.test_stt::test_recognize[livekit.plugins.gradium.STT]	livekit.plugins.gradium.STT: Gradium API key is required. Pass one in via the api_key parameter, or set it as the GRADIUM_API_KEY environment variable
tests.test_stt::test_recognize[livekit.plugins.aws]	unknown@Amazon Transcribe does not support batch recognition
tests.test_stt::test_stream[livekit.plugins.deepgram]	livekit.plugins.deepgram: Deepgram API key is required, either as argument or set DEEPGRAM_API_KEY environment variable
tests.test_stt::test_recognize[livekit.plugins.sarvam]	livekit.plugins.sarvam: Sarvam API key is required. Provide it directly or set SARVAM_API_KEY environment variable.
tests.test_stt::test_stream[livekit.plugins.assemblyai]	livekit.plugins.assemblyai: AssemblyAI API key is required. Pass one in via the api_key parameter, or set it as the ASSEMBLYAI_API_KEY environment variable
tests.test_stt::test_recognize[livekit.plugins.cartesia]	livekit.plugins.cartesia: Cartesia API key is required, either as argument or set CARTESIA_API_KEY environment variable
tests.test_stt::test_recognize[livekit.plugins.cartesia._legacy]	livekit.plugins.cartesia._legacy: Cartesia API key is required, either as argument or set CARTESIA_API_KEY environment variable
tests.test_stt::test_stream[livekit.plugins.speechmatics]	livekit.plugins.speechmatics: Missing Speechmatics API key
tests.test_stt::test_recognize[livekit.plugins.soniox]	livekit.plugins.soniox: Soniox API key is required. Set SONIOX_API_KEY or pass api_key
...	... and 20 more

---

Triggered by workflow run #1

[grace0950]

關閉：這張是舊的 pathors → main。fork 的慣例是 main 只當 upstream 鏡像、我們的改動都在 pathors，不會 merge 回 main。pathors 已於 2026-10-29 升到 livekit-agents 1.8.9（pathorsAI/pathors#4095）。