chore: sync sdk code with DeepLearning repo

assemblyai/__version__.py

@@ -1 +1 @@
-__version__ = "0.54.0"
+__version__ = "0.54.1"

assemblyai/streaming/v3/__init__.py

@@ -5,6 +5,7 @@
     EventMessage,
     LLMGatewayResponseEvent,
     SpeechModel,
+    SpeechStartedEvent,
     StreamingClientOptions,
     StreamingError,
     StreamingEvents,
@@ -21,6 +22,7 @@
     "EventMessage",
     "LLMGatewayResponseEvent",
     "SpeechModel",
+    "SpeechStartedEvent",
     "StreamingClient",
     "StreamingClientOptions",
     "StreamingError",

assemblyai/streaming/v3/client.py

@@ -20,6 +20,7 @@
     ForceEndpoint,
     LLMGatewayResponseEvent,
     OperationMessage,
+    SpeechStartedEvent,
     StreamingClientOptions,
     StreamingError,
     StreamingErrorCodes,
@@ -236,6 +237,8 @@ def _parse_message(self, data: Dict[str, Any]) -> Optional[EventMessage]:
                 return TerminationEvent.model_validate(data)
             elif event_type == StreamingEvents.Turn:
                 return TurnEvent.model_validate(data)
+            elif event_type == StreamingEvents.SpeechStarted:
+                return SpeechStartedEvent.model_validate(data)
             elif event_type == StreamingEvents.LLMGatewayResponse:
                 return LLMGatewayResponseEvent.model_validate(data)
             else:

assemblyai/streaming/v3/models.py

@@ -49,6 +49,11 @@ class TerminationEvent(BaseModel):
     session_duration_seconds: Optional[int] = None
 
 
+class SpeechStartedEvent(BaseModel):
+    type: Literal["SpeechStarted"] = "SpeechStarted"
+    timestamp: int
+
+
 class ErrorEvent(BaseModel):
     error: str
 
@@ -64,6 +69,7 @@ class LLMGatewayResponseEvent(BaseModel):
     BeginEvent,
     TerminationEvent,
     TurnEvent,
+    SpeechStartedEvent,
     ErrorEvent,
     LLMGatewayResponseEvent,
 ]
@@ -175,5 +181,6 @@ class StreamingEvents(Enum):
     Begin = "Begin"
     Termination = "Termination"
     Turn = "Turn"
+    SpeechStarted = "SpeechStarted"
     Error = "Error"
     LLMGatewayResponse = "LLMGatewayResponse"

tests/unit/test_streaming.py

@@ -4,6 +4,7 @@
 
 from assemblyai.streaming.v3 import (
     SpeechModel,
+    SpeechStartedEvent,
     StreamingClient,
     StreamingClientOptions,
     StreamingParameters,
@@ -352,3 +353,14 @@ def test_turn_event_without_speaker_label():
     }
     event = TurnEvent.parse_obj(data)
     assert event.speaker_label is None
+
+
+def test_speech_started_event():
+    """Test SpeechStarted event parsing (u3-rt-pro only)"""
+    data = {
+        "type": "SpeechStarted",
+        "timestamp": 1280,
+    }
+    event = SpeechStartedEvent.parse_obj(data)
+    assert event.type == "SpeechStarted"
+    assert event.timestamp == 1280