make structured output

backend/agents/ext_agents.py

@@ -1,45 +1,33 @@
 from typing import List, Optional
 
 from autogen_core.base import MessageContext
-from autogen_core.components import (DefaultTopicId, RoutedAgent,
-                                     default_subscription, message_handler,
-                                     type_subscription)
+from autogen_core.components import (
+    DefaultTopicId,
+    RoutedAgent,
+    default_subscription,
+    message_handler,
+    type_subscription,
+)
 from llama_index.core.agent.runner.base import AgentRunner
 from llama_index.core.base.llms.types import ChatMessage, MessageRole
 from llama_index.core.chat_engine.types import AgentChatResponse
 from llama_index.core.memory.types import BaseMemory
-from pydantic import BaseModel
 
-from ..data_types import AgentResponse, EndUserMessage
+from ..data_types import AgentStructuredResponse, EndUserMessage, Resource
 from ..otlp_tracing import logger
 
 
-class Resource(BaseModel):
-    """
-    Represents a resource node retrieved during chat interactions.
-
-    Attributes:
-        content (str): The textual content of the resource.
-        node_id (str): The identifier of the node.
-        score (Optional[float]): Score representing the relevance of the resource.
-    """
-
-    content: str
-    node_id: str
-    score: Optional[float] = None
+# class Message(BaseModel):
+#     """
+#     Represents a message exchanged during the chat.
 
+#     Attributes:
+#         content (str): The textual content of the message.
+#         sources (Optional[List[Resource]]): List of resources associated with the message.
+#     """
 
-class Message(BaseModel):
-    """
-    Represents a message exchanged during the chat.
-
-    Attributes:
-        content (str): The textual content of the message.
-        sources (Optional[List[Resource]]): List of resources associated with the message.
-    """
-
-    content: str
-    sources: Optional[List[Resource]] = None
+#     content: str
+#     sources: Optional[List[Resource]] = None
 
 
 @default_subscription
@@ -116,17 +104,19 @@ async def handle_user_message(
             resources.extend(tools)
             logger.info(response.response)
             await self.publish_message(
-                AgentResponse(
-                    source="LlamaIndexAgent",
-                    content=f"\n{response.response}\n",
+                AgentStructuredResponse(
+                    agent_type="default_agent",
+                    data=None,
+                    message=f"\n{response.response}\n",
                 ),
                 DefaultTopicId(type="user_proxy", source=self._session_id),
             )
         else:
             await self.publish_message(
-                AgentResponse(
-                    source="LlamaIndexAgent",
-                    content="I'm sorry, I don't have an answer for you.",
+                AgentStructuredResponse(
+                    agent_type="default_agent",
+                    data=None,
+                    message="I'm sorry, I don't have an answer for you.",
                 ),
                 DefaultTopicId(type="user_proxy", source=self._session_id),
             )

backend/agents/travel_activities.py

@@ -26,6 +26,7 @@
     GroupChatMessage,
     HandoffMessage,
     TravelRequest,
+    AgentStructuredResponse,
 )
 from ..otlp_tracing import logger
 
@@ -171,9 +172,10 @@ async def handle_message(
 
         # Publish the response to the group chat manager
         await self.publish_message(
-            AgentResponse(
-                source=self.id.type,
-                content=activities_structured.model_dump_json(),
+            AgentStructuredResponse(
+                agent_type=self.id.type,
+                data=activities_structured,
+                message=f"Activities processed successfully for query - {message.content}",
             ),
             DefaultTopicId(type="user_proxy", source=ctx.topic_id.source),
         )

backend/agents/travel_car.py

@@ -1,8 +1,8 @@
 import asyncio
 import datetime
 import random
-from typing import Dict, List
-
+from typing import List
+from autogen_core.components.tools import FunctionTool, Tool
 from autogen_core.base import MessageContext
 from autogen_core.components import (
     DefaultTopicId,
@@ -12,13 +12,13 @@
 )
 from autogen_core.components.models import LLMMessage, SystemMessage
 from typing_extensions import Annotated
-
 from ..data_types import (
-    AgentResponse,
+    AgentStructuredResponse,
     EndUserMessage,
     GroupChatMessage,
     HandoffMessage,
     TravelRequest,
+    CarRental,
 )
 from ..otlp_tracing import logger
 
@@ -31,33 +31,7 @@ async def simulate_car_rental_booking(
     rental_end_date: Annotated[
         str, "The end date of the car rental in the format 'YYYY-MM-DD'."
     ],
-) -> Dict[str, str | int]:
-    """
-    Simulate a car rental booking process.
-
-    This function simulates the process of booking a car rental by randomly selecting a car option,
-    calculating the rental duration and total price, and generating a booking reference number.
-
-    Args:
-        rental_city (str): The city where the car rental will take place.
-        rental_start_date (str): The start date of the car rental in the format 'YYYY-MM-DD'.
-        rental_end_date (str): The end date of the car rental in the format 'YYYY-MM-DD'.
-
-    Returns:
-        Dict[str, str | int]: A dictionary containing the car rental details, including the rental city,
-                              start and end dates, car type, company, total price, and booking reference.
-                              Example:
-                              {
-                                  "rental_city": "New York",
-                                  "rental_start_date": "2023-10-01",
-                                  "rental_end_date": "2023-10-07",
-                                  "car_type": "SUV",
-                                  "company": "Hertz",
-                                  "total_price": 560,
-                                  "booking_reference": "CR-1234-NYC"
-                              }
-    """
-    # Simulate available car options
+) -> CarRental:
     car_options = [
         {"car_type": "Sedan", "company": "Avis", "price_per_day": 50},
         {"car_type": "SUV", "company": "Hertz", "price_per_day": 80},
@@ -76,36 +50,37 @@ async def simulate_car_rental_booking(
         {"car_type": "Hatchback", "company": "City Rentals", "price_per_day": 45},
     ]
 
-    # Randomly select a car option
     selected_car = random.choice(car_options)
-
-    # Calculate rental duration
     start_date = datetime.datetime.strptime(rental_start_date, "%Y-%m-%d")
     end_date = datetime.datetime.strptime(rental_end_date, "%Y-%m-%d")
     rental_days = (end_date - start_date).days
-
-    # Calculate total price
     total_price = rental_days * selected_car["price_per_day"]
-
-    # Create a booking reference number
     booking_reference = f"CR-{random.randint(1000, 9999)}-{rental_city[:3].upper()}"
 
-    # Simulate car rental details
-    car_rental_details = {
-        "rental_city": rental_city,
-        "rental_start_date": rental_start_date,
-        "rental_end_date": rental_end_date,
-        "car_type": selected_car["car_type"],
-        "company": selected_car["company"],
-        "total_price": total_price,
-        "booking_reference": booking_reference,
-    }
-    # Induce an artificial delay to simulate network latency
-    await asyncio.sleep(3)
+    car_rental_details = CarRental(
+        rental_city=rental_city,
+        rental_start_date=rental_start_date,
+        rental_end_date=rental_end_date,
+        car_type=selected_car["car_type"],
+        company=selected_car["company"],
+        total_price=total_price,
+        booking_reference=booking_reference,
+    )
+
+    await asyncio.sleep(2)
     return car_rental_details
 
 
-# Car Rental Agent
+def get_car_rental_tool() -> List[Tool]:
+    return [
+        FunctionTool(
+            name="simulate_car_rental_booking",
+            func=simulate_car_rental_booking,
+            description="Simulates a car rental booking process based on user preferences.",
+        )
+    ]
+
+
 @type_subscription("car_rental")
 class CarRentalAgent(RoutedAgent):
     def __init__(self) -> None:
@@ -117,32 +92,35 @@ def __init__(self) -> None:
             )
         ]
 
-    async def _process_request(self, requirements: dict) -> dict:
-        # Simulate car rental booking logic
-        return await simulate_car_rental_booking(
-            requirements.get("rental_city", "Unknown"),
-            requirements.get("rental_start_date", "Unknown"),
-            requirements.get("rental_end_date", "Unknown"),
-        )
-
     @message_handler
     async def handle_message(
         self, message: EndUserMessage, ctx: MessageContext
     ) -> None:
         logger.info(f"CarRentalAgent received message: {message.content}")
         if "travel plan" in message.content.lower():
-            # Cannot handle complex travel plans, hand off back to router
             await self.publish_message(
                 HandoffMessage(content=message.content, source=self.id.type),
                 DefaultTopicId(type="router", source=ctx.topic_id.source),
             )
             return
-        # Extract requirements and process the car rental request
-        requirements = self.extract_requirements(message.content)
-        response = await self._process_request(requirements)
+
+        # You would typically call a LLM to extract the requirement or have a function call here
+        requirements = {
+            "rental_city": (
+                "New York" if "new york" in message.content.lower() else "Unknown"
+            ),
+            "rental_start_date": "2023-12-21",
+            "rental_end_date": "2023-12-26",
+        }
+        response = await simulate_car_rental_booking(
+            requirements["rental_city"],
+            requirements["rental_start_date"],
+            requirements["rental_end_date"],
+        )
         await self.publish_message(
-            AgentResponse(
-                source=self.id.type,
+            AgentStructuredResponse(
+                agent_type=self.id.type,
+                data=response,
                 content=f"Car rented: {response}",
             ),
             DefaultTopicId(type="user_proxy", source=ctx.topic_id.source),
@@ -155,18 +133,19 @@ async def handle_travel_request(
         logger.info(
             f"CarRentalAgent received travel request: TravelRequest - {message.content}"
         )
-        requirements = self.extract_requirements(message.content)
-        response = await self._process_request(requirements)
+        requirements = {
+            "rental_city": (
+                "New York" if "new york" in message.content.lower() else "Unknown"
+            ),
+            "rental_start_date": "2023-12-21",
+            "rental_end_date": "2023-12-26",
+        }
+        response = await simulate_car_rental_booking(
+            requirements["rental_city"],
+            requirements["rental_start_date"],
+            requirements["rental_end_date"],
+        )
         return GroupChatMessage(
             source=self.id.type,
             content=f"Car rented: {response}",
         )
-
-    def extract_requirements(self, user_input: str) -> dict:
-        # Simple keyword-based extraction
-        requirements = {}
-        if "new york" in user_input.lower():
-            requirements["rental_city"] = "New York"
-        requirements["rental_start_date"] = "2023-12-21"
-        requirements["rental_end_date"] = "2023-12-26"
-        return requirements

backend/agents/travel_destination.py

@@ -2,14 +2,22 @@
 from typing import List
 
 from autogen_core.base import MessageContext
-from autogen_core.components import (DefaultTopicId, RoutedAgent,
-                                     message_handler, type_subscription)
-from autogen_core.components.models import (LLMMessage, SystemMessage,
-                                            UserMessage)
+from autogen_core.components import (
+    DefaultTopicId,
+    RoutedAgent,
+    message_handler,
+    type_subscription,
+)
+from autogen_core.components.models import LLMMessage, SystemMessage, UserMessage
 from autogen_ext.models import AzureOpenAIChatCompletionClient
 
-from ..data_types import (AgentResponse, DestinationInfo, EndUserMessage,
-                          GroupChatMessage, TravelRequest)
+from ..data_types import (
+    AgentStructuredResponse,
+    DestinationInfo,
+    EndUserMessage,
+    GroupChatMessage,
+    TravelRequest,
+)
 from ..otlp_tracing import logger
 
 
@@ -33,7 +41,7 @@ async def handle_message(
         self, message: EndUserMessage, ctx: MessageContext
     ) -> None:
         logger.info(
-            f"DestinationAgent received travel request: EndUserMessage{message.content}"
+            f"DestinationAgent received travel request: EndUserMessage {message.content}"
         )
         # Provide destination information
         try:
@@ -55,9 +63,10 @@ async def handle_message(
             pass
 
         await self.publish_message(
-            AgentResponse(
-                source=self.id.type,
-                content=destination_info_structured.model_dump_json(),
+            AgentStructuredResponse(
+                agent_type=self.id.type,
+                data=destination_info_structured,
+                message=message.content,
             ),
             DefaultTopicId(type="user_proxy", source=ctx.topic_id.source),
         )