Initial fix

playground/streaming/agent/chat.py

@@ -93,7 +93,7 @@ async def receiver():
                     agent_response = typing.cast(AgentResponseMessage, response)
 
                     agent.transcript.add_bot_message(
-                        agent_response.message.text, conversation_id
+                        agent_response.message.text, conversation_id, agent_response.message_id
                     )
                     print(
                         "AI: "

vocode/streaming/agent/base_agent.py

@@ -16,6 +16,7 @@
     TYPE_CHECKING,
 )
 import typing
+import uuid
 from opentelemetry import trace
 from opentelemetry.trace import Span
 from vocode.streaming.action.factory import ActionFactory
@@ -90,6 +91,7 @@ class AgentResponse(TypedModel, type=AgentResponseType.BASE.value):
 
 
 class AgentResponseMessage(AgentResponse, type=AgentResponseType.MESSAGE.value):
+    message_id: str
     message: BaseMessage
     is_interruptible: bool = True
 
@@ -180,7 +182,7 @@ def start(self):
         super().start()
         if self.agent_config.initial_message is not None:
             self.produce_interruptible_agent_response_event_nonblocking(
-                AgentResponseMessage(message=self.agent_config.initial_message),
+                AgentResponseMessage(message=self.agent_config.initial_message, message_id="initial_message"),
                 is_interruptible=False,
             )
 
@@ -214,6 +216,7 @@ async def handle_generate_response(
         agent_span_first = tracer.start_span(
             f"{tracer_name_start}.generate_first"  # type: ignore
         )
+        response_id = str(uuid.uuid4())
         responses = self.generate_response(
             transcription.message,
             is_interrupt=transcription.is_interrupt,
@@ -229,7 +232,7 @@ async def handle_generate_response(
                 agent_span_first.end()
                 is_first_response = False
             self.produce_interruptible_agent_response_event_nonblocking(
-                AgentResponseMessage(message=BaseMessage(text=response)),
+                AgentResponseMessage(message_id=response_id, message=BaseMessage(text=response)),
                 is_interruptible=self.agent_config.allow_agent_to_be_cut_off,
             )
         # TODO: implement should_stop for generate_responses

vocode/streaming/models/transcript.py

@@ -17,6 +17,7 @@ def to_string(self, include_timestamp: bool = False) -> str:
 
 
 class Message(EventLog):
+    message_id: Optional[str]
     text: str
 
     def to_string(self, include_timestamp: bool = False) -> str:
@@ -73,9 +74,22 @@ def add_message(
         text: str,
         sender: Sender,
         conversation_id: str,
+        message_id: Optional[str] = None,
     ):
-        timestamp = time.time()
-        self.event_logs.append(Message(text=text, sender=sender, timestamp=timestamp))
+        if (
+            self.event_logs
+            and (last_event := self.event_logs[-1]).sender == Sender.BOT
+            and message_id
+            and last_event.message_id == message_id
+        ):
+            last_event.text += f" {text}"
+        else:
+            timestamp = time.time()
+            self.event_logs.append(
+                Message(
+                    text=text, sender=sender, timestamp=timestamp, message_id=message_id
+                )
+            )
         if self.events_manager is not None:
             self.events_manager.publish_event(
                 TranscriptEvent(
@@ -93,11 +107,12 @@ def add_human_message(self, text: str, conversation_id: str):
             conversation_id=conversation_id,
         )
 
-    def add_bot_message(self, text: str, conversation_id: str):
+    def add_bot_message(self, text: str, conversation_id: str, message_id: str):
         self.add_message(
             text=text,
             sender=Sender.BOT,
             conversation_id=conversation_id,
+            message_id=message_id,
         )
 
     def get_last_user_message(self):

vocode/streaming/streaming_conversation.py

@@ -274,6 +274,7 @@ async def process(self, item: InterruptibleAgentResponseEvent[AgentResponse]):
                     self.conversation.transcript.add_bot_message(
                         text=agent_response.message.text,
                         conversation_id=self.conversation.id,
+                        message_id=agent_response.message_id,
                     )
 
                 agent_response_message = typing.cast(