Enhance pipeline audio in the executor

enhance_chunk blocks the event loop for a bit every
time its called because its CPU intensive

homeassistant/components/assist_pipeline/pipeline.py

@@ -575,8 +575,11 @@ class PipelineRun:
     _device_id: str | None = None
     """Optional device id set during run start."""
 
+    _loop: asyncio.AbstractEventLoop | None = field(init=False, default=None)
+
     def __post_init__(self) -> None:
         """Set language for pipeline."""
+        self._loop = asyncio.get_running_loop()
         self.language = self.pipeline.language or self.hass.config.language
 
         # wake -> stt -> intent -> tts
@@ -1284,6 +1287,7 @@ class PipelineRun:
         assert self.audio_enhancer is not None
 
         timestamp_ms = 0
+        assert self._loop is not None
         async for dirty_samples in audio_stream:
             if self.audio_settings.volume_multiplier != 1.0:
                 # Static gain
@@ -1295,7 +1299,9 @@ class PipelineRun:
             for dirty_chunk in chunk_samples(
                 dirty_samples, BYTES_PER_CHUNK, self.audio_chunking_buffer
             ):
-                yield self.audio_enhancer.enhance_chunk(dirty_chunk, timestamp_ms)
+                yield await self._loop.run_in_executor(
+                    None, self.audio_enhancer.enhance_chunk, dirty_chunk, timestamp_ms
+                )
                 timestamp_ms += MS_PER_CHUNK
 
 

tests/components/assist_pipeline/test_init.py

@@ -708,7 +708,7 @@ async def test_wake_word_detection_aborted(
     assert process_events(events) == snapshot
 
 
-def test_pipeline_run_equality(hass: HomeAssistant, init_components) -> None:
+async def test_pipeline_run_equality(hass: HomeAssistant, init_components) -> None:
     """Test that pipeline run equality uses unique id."""
 
     def event_callback(event):