configurable cut off response

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "vocode"
-version = "0.1.34"
+version = "0.1.35"
 description = "The all-in-one voice SDK"
 authors = ["Ajay Raj <ajay@vocode.dev>"]
 license = "MIT License"

simple_conversation.py

@@ -19,6 +19,7 @@ from vocode.models.transcriber import (
 )
 from vocode.models.agent import (
     ChatGPTAgentConfig,
+    CutOffResponse,
     FillerAudioConfig,
     RESTfulUserImplementedAgentConfig,
     WebSocketUserImplementedAgentConfig,
@@ -46,12 +47,11 @@ if __name__ == "__main__":
         transcriber_config=DeepgramTranscriberConfig.from_input_device(
             microphone_input
         ),
-        agent_config=WebSocketUserImplementedAgentConfig(
+        agent_config=ChatGPTAgentConfig(
             initial_message=BaseMessage(text="Hello!"),
+            prompt_preamble="The AI is having a pleasant conversation about life",
             generate_responses=True,
-            respond=WebSocketUserImplementedAgentConfig.RouteConfig(
-                url="wss://9b1ff0eee874.ngrok.app/respond",
-            ),
+            cut_off_response=CutOffResponse(),
         ),
         synthesizer_config=AzureSynthesizerConfig.from_output_device(speaker_output),
     )

vocode/models/agent.py

@@ -45,12 +45,17 @@ class AgentConfig(TypedModel, type=AgentType.BASE):
     send_filler_audio: Union[bool, FillerAudioConfig] = False
 
 
+class CutOffResponse(BaseModel):
+    messages: list[BaseMessage] = [BaseMessage(text="Sorry?")]
+
+
 class LLMAgentConfig(AgentConfig, type=AgentType.LLM):
     prompt_preamble: str
     expected_first_prompt: Optional[str] = None
     model_name: str = LLM_AGENT_DEFAULT_MODEL_NAME
     temperature: float = LLM_AGENT_DEFAULT_TEMPERATURE
     max_tokens: int = LLM_AGENT_DEFAULT_MAX_TOKENS
+    cut_off_response: Optional[CutOffResponse] = None
 
 
 class ChatGPTAlphaAgentConfig(AgentConfig, type=AgentType.CHAT_GPT_ALPHA):
@@ -66,6 +71,7 @@ class ChatGPTAgentConfig(AgentConfig, type=AgentType.CHAT_GPT):
     generate_responses: bool = False
     temperature: float = LLM_AGENT_DEFAULT_TEMPERATURE
     max_tokens: int = LLM_AGENT_DEFAULT_MAX_TOKENS
+    cut_off_response: Optional[CutOffResponse] = None
 
 
 class InformationRetrievalAgentConfig(