Merge pull request #122 from GetStream/cleanup_stt

Cleanup STT

Author: tschellenbach

agents-core/vision_agents/core/agents/agents.py

@@ -177,6 +177,7 @@ async def simple_response(
         """
         Overwrite simple_response if you want to change how the Agent class calls the LLM
         """
+        logger.info("asking LLM to reply to %s", text)
         with self.tracer.start_as_current_span("simple_response") as span:
             response = await self.llm.simple_response(
                 text=text, processors=self.processors, participant=participant

agents-core/vision_agents/core/stt/__init__.py

@@ -1,3 +1,4 @@
 from .stt import STT
+from .events import TranscriptResponse
 
-__all__ = ["STT"]
+__all__ = ["STT", "TranscriptResponse"]

agents-core/vision_agents/core/stt/events.py

@@ -4,22 +4,47 @@
 
 
 @dataclass
-class STTTranscriptEvent(PluginBaseEvent):
-    """Event emitted when a complete transcript is available."""
-
-    type: str = field(default='plugin.stt_transcript', init=False)
-    text: str = ""
+class TranscriptResponse:
     confidence: Optional[float] = None
     language: Optional[str] = None
     processing_time_ms: Optional[float] = None
     audio_duration_ms: Optional[float] = None
     model_name: Optional[str] = None
-    words: Optional[list[dict[str, Any]]] = None
+    other: Optional[dict] = None
+
+@dataclass
+class STTTranscriptEvent(PluginBaseEvent):
+    """Event emitted when a complete transcript is available."""
+
+    type: str = field(default='plugin.stt_transcript', init=False)
+    text: str = ""
+    response: TranscriptResponse = field(default_factory=TranscriptResponse)
     is_final: bool = True
 
     def __post_init__(self):
         if not self.text:
             raise ValueError("Transcript text cannot be empty")
+    
+    # Convenience properties for backward compatibility
+    @property
+    def confidence(self) -> Optional[float]:
+        return self.response.confidence
+    
+    @property
+    def language(self) -> Optional[str]:
+        return self.response.language
+    
+    @property
+    def processing_time_ms(self) -> Optional[float]:
+        return self.response.processing_time_ms
+    
+    @property
+    def audio_duration_ms(self) -> Optional[float]:
+        return self.response.audio_duration_ms
+    
+    @property
+    def model_name(self) -> Optional[str]:
+        return self.response.model_name
 
 
 @dataclass
@@ -28,13 +53,29 @@ class STTPartialTranscriptEvent(PluginBaseEvent):
 
     type: str = field(default='plugin.stt_partial_transcript', init=False)
     text: str = ""
-    confidence: Optional[float] = None
-    language: Optional[str] = None
-    processing_time_ms: Optional[float] = None
-    audio_duration_ms: Optional[float] = None
-    model_name: Optional[str] = None
-    words: Optional[list[dict[str, Any]]] = None
+    response: TranscriptResponse = field(default_factory=TranscriptResponse)
     is_final: bool = False
+    
+    # Convenience properties for backward compatibility
+    @property
+    def confidence(self) -> Optional[float]:
+        return self.response.confidence
+    
+    @property
+    def language(self) -> Optional[str]:
+        return self.response.language
+    
+    @property
+    def processing_time_ms(self) -> Optional[float]:
+        return self.response.processing_time_ms
+    
+    @property
+    def audio_duration_ms(self) -> Optional[float]:
+        return self.response.audio_duration_ms
+    
+    @property
+    def model_name(self) -> Optional[str]:
+        return self.response.model_name
 
 
 @dataclass