Fix voice-agent backend type

Author: ajhai

llmstack/apps/runner/agent_controller.py

@@ -5,7 +5,7 @@
 import queue
 import ssl
 import threading
-from typing import Any, Dict, List, Literal, Optional, Union
+from typing import Any, Dict, List, Optional, Union
 
 import websockets
 from asgiref.sync import sync_to_async
@@ -370,7 +370,7 @@ def process(self, data: AgentControllerData):
             )
 
     async def process_messages(self, data: AgentControllerData):
-        if self._config.is_voice_agent and self._config.agent_config.backend.backend_type == Literal["multi_modal"]:
+        if self._config.is_voice_agent and self._config.agent_config.backend.backend_type == "multi_modal":
             if not self._websocket:
                 await self._init_websocket_connection()
 

llmstack/apps/runner/app_coordinator.py

@@ -55,7 +55,7 @@ def __init__(
         }
 
         # Add agent actor for output if it's an agent
-        if self._is_agent:
+        if self._is_agent or self._is_voice_agent:
             self.actors["output"] = AgentActor.start(
                 coordinator_urn=self.actor_urn,
                 dependencies=["_inputs0"] + list(self._actor_configs_map.keys()),

llmstack/apps/types/voice_agent.py

@@ -66,7 +66,7 @@ class VoiceActivityDetectionSettings(BaseSchema):
 
 class MultiModal(BaseSchema):
     backend_type: Literal["multi_modal"] = Field(
-        default=Literal["multi_modal"],
+        default="multi_modal",
         title="Backend Type",
         description="The backend type to use for the multi modal model.",
         json_schema_extra={"widget": "hidden", "readOnly": True},
@@ -113,7 +113,7 @@ class MultiModal(BaseSchema):
 
 class CustomPipeline(BaseSchema):
     backend_type: Literal["custom_pipeline"] = Field(
-        default=Literal["custom_pipeline"],
+        default="custom_pipeline",
         title="Backend Type",
         description="The backend type to use for the custom pipeline.",
         json_schema_extra={"widget": "hidden", "readOnly": True},
@@ -190,7 +190,7 @@ class VoiceAgentConfigSchema(BaseSchema):
         ge=1,
     )
     backend: Union[MultiModal, CustomPipeline] = Field(
-        default=None,
+        default=MultiModal(),
         title="Backend",
         description="The backend to use for the voice agent.",
     )