Publish user SST text

Author: kPatch

coffee_ws/src/coffee_voice_agent/coffee_voice_agent/voice_agent_bridge.py

@@ -70,6 +70,13 @@ def __init__(self):
             callback_group=self.callback_group
         )
 
+        self.user_speech_pub = self.create_publisher(
+            String,
+            'voice_agent/user_speech',
+            10,
+            callback_group=self.callback_group
+        )
+        
         self.connected_pub = self.create_publisher(
             Bool, 
             'voice_agent/connected', 
@@ -216,6 +223,18 @@ async def _handle_websocket_message(self, message: str):
 
                 self.tool_event_pub.publish(tool_msg)
 
+            elif message_type == 'USER_SPEECH':
+                # Handle user speech transcription events
+                speech_data = data.get('data', {})
+                user_text = speech_data.get('text', '')
+                
+                self.get_logger().info(f"User Speech: '{user_text[:50]}{'...' if len(user_text) > 50 else ''}'")
+                
+                # Publish user speech to ROS2 topic
+                speech_msg = String()
+                speech_msg.data = user_text
+                self.user_speech_pub.publish(speech_msg)
+                
             elif message_type == 'ACKNOWLEDGMENT':
                 # Handle acknowledgment messages from voice agent
                 status = data.get('status', 'unknown')

coffee_ws/src/coffee_voice_agent/scripts/state/state_manager.py

@@ -228,6 +228,9 @@ async def handle_conversation_item():
                             text_lower = user_text.lower()
                             logger.info(f"🔍 DEBUG: User said: '{user_text}'")
 
+                            # Send user speech event
+                            await self._send_user_speech_event(user_text)
+                            
                             goodbye_words = ['goodbye', 'thanks', 'that\'s all', 'see you', 'bye']
 
                             if any(word in text_lower for word in goodbye_words):
@@ -622,4 +625,15 @@ async def _send_tool_event(self, tool_name: str, status: str, parameters: list =
             }
             await self.agent._send_websocket_event("TOOL_EVENT", tool_data)
         else:
-            logger.debug(f"Cannot send tool event - no agent WebSocket connection") 
+            logger.debug(f"Cannot send tool event - no agent WebSocket connection")
+
+    async def _send_user_speech_event(self, text: str):
+        """Send user speech event through agent's WebSocket connection"""
+        if self.agent and hasattr(self.agent, '_send_websocket_event'):
+            speech_data = {
+                "text": text,
+                "timestamp": datetime.now().isoformat()
+            }
+            await self.agent._send_websocket_event("USER_SPEECH", speech_data)
+        else:
+            logger.debug(f"Cannot send user speech event - no agent WebSocket connection")