Python: [Breaking] Python: Respond with AgentRunResponse with serialized structured output

python/packages/azurefunctions/agent_framework_azurefunctions/_entities.py

@@ -9,9 +9,7 @@
 
 import asyncio
 import inspect
-import json
 from collections.abc import AsyncIterable, Callable
-from datetime import datetime, timezone
 from typing import Any, cast
 
 import azure.durable_functions as df
@@ -30,11 +28,10 @@
     DurableAgentState,
     DurableAgentStateData,
     DurableAgentStateEntry,
-    DurableAgentStateMessage,
     DurableAgentStateRequest,
     DurableAgentStateResponse,
 )
-from ._models import AgentResponse, RunRequest
+from ._models import RunRequest
 
 logger = get_logger("agent_framework.azurefunctions.entities")
 
@@ -97,21 +94,16 @@ async def run_agent(
         self,
         context: df.DurableEntityContext,
         request: RunRequest | dict[str, Any] | str,
-    ) -> dict[str, Any]:
+    ) -> AgentRunResponse:
         """Execute the agent with a message directly in the entity.
 
         Args:
             context: Entity context
             request: RunRequest object, dict, or string message (for backward compatibility)
 
         Returns:
-            Dict with status information and response (serialized AgentResponse)
-
-        Note:
-            The agent returns an AgentRunResponse object which is stored in state.
-            This method extracts the text/structured response and returns an AgentResponse dict.
+            AgentRunResponse enriched with execution metadata.
         """
-        # Convert string or dict to RunRequest
         if isinstance(request, str):
             run_request = RunRequest(message=request, role=Role.USER)
         elif isinstance(request, dict):
@@ -135,8 +127,6 @@ async def run_agent(
         logger.debug(f"[AgentEntity.run_agent] Received Message: {state_request}")
 
         try:
-            logger.debug("[AgentEntity.run_agent] Starting agent invocation")
-
             # Build messages from conversation history, excluding error responses
             # Error responses are kept in history for tracking but not sent to the agent
             chat_messages: list[ChatMessage] = [
@@ -164,83 +154,42 @@ async def run_agent(
                 type(agent_run_response).__name__,
             )
 
-            response_text = None
-            structured_response = None
-            response_str: str | None = None
-
             try:
-                if response_format:
-                    try:
-                        response_str = agent_run_response.text
-                        structured_response = json.loads(response_str)
-                        logger.debug("Parsed structured JSON response")
-                    except json.JSONDecodeError as decode_error:
-                        logger.warning(f"Failed to parse JSON response: {decode_error}")
-                        response_text = response_str
-                else:
-                    raw_text = agent_run_response.text
-                    response_text = raw_text if raw_text else "No response"
-                    preview = response_text
-                    logger.debug(f"Response: {preview[:100]}..." if len(preview) > 100 else f"Response: {preview}")
+                response_text = agent_run_response.text if agent_run_response.text else "No response"
+                logger.debug(f"Response: {response_text[:100]}...")
             except Exception as extraction_error:
                 logger.error(
-                    f"Error extracting response: {extraction_error}",
+                    "Error extracting response text: %s",
+                    extraction_error,
                     exc_info=True,
                 )
-                response_text = "Error extracting response"
 
             state_response = DurableAgentStateResponse.from_run_response(correlation_id, agent_run_response)
             self.state.data.conversation_history.append(state_response)
 
-            agent_response = AgentResponse(
-                response=response_text,
-                message=str(message),
-                thread_id=str(thread_id),
-                status="success",
-                message_count=len(self.state.data.conversation_history),
-                structured_response=structured_response,
-            )
-            result = agent_response.to_dict()
-
             logger.debug("[AgentEntity.run_agent] AgentRunResponse stored in conversation history")
 
-            return result
+            return agent_run_response
 
         except Exception as exc:
             import traceback
 
             error_traceback = traceback.format_exc()
-            logger.error("[AgentEntity.run_agent] Agent execution failed")
-            logger.error(f"Error: {exc!s}")
-            logger.error(f"Error type: {type(exc).__name__}")
-            logger.error(f"Full traceback:\n{error_traceback}")
+            logger.error(f"[AgentEntity.run_agent] Agent execution failed. Full traceback:\n{error_traceback}")
 
             # Create error message
-            error_message = DurableAgentStateMessage.from_chat_message(
-                ChatMessage(
-                    role=Role.ASSISTANT, contents=[ErrorContent(message=str(exc), error_code=type(exc).__name__)]
-                )
+            error_message = ChatMessage(
+                role=Role.ASSISTANT, contents=[ErrorContent(message=str(exc), error_code=type(exc).__name__)]
             )
 
+            error_response = AgentRunResponse(messages=[error_message])
+
             # Create and store error response in conversation history
-            error_state_response = DurableAgentStateResponse(
-                correlation_id=correlation_id,
-                created_at=datetime.now(tz=timezone.utc),
-                messages=[error_message],
-                is_error=True,
-            )
+            error_state_response = DurableAgentStateResponse.from_run_response(correlation_id, error_response)
+            error_state_response.is_error = True
             self.state.data.conversation_history.append(error_state_response)
 
-            error_response = AgentResponse(
-                response=f"Error: {exc!s}",
-                message=str(message),
-                thread_id=str(thread_id),
-                status="error",
-                message_count=len(self.state.data.conversation_history),
-                error=str(exc),
-                error_type=type(exc).__name__,
-            )
-            return error_response.to_dict()
+            return error_response
 
     async def _invoke_agent(
         self,
@@ -432,7 +381,7 @@ async def _entity_coroutine(context: df.DurableEntityContext) -> None:
                     request = "" if input_data is None else str(cast(object, input_data))
 
                 result = await entity.run_agent(context, request)
-                context.set_result(result)
+                context.set_result(result.to_dict())
 
             elif operation == "reset":
                 entity.reset(context)
@@ -442,8 +391,9 @@ async def _entity_coroutine(context: df.DurableEntityContext) -> None:
                 logger.error("[entity_function] Unknown operation: %s", operation)
                 context.set_result({"error": f"Unknown operation: {operation}"})
 
-            logger.debug("State dict: %s", entity.state.to_dict())
-            context.set_state(entity.state.to_dict())
+            serialized_state = entity.state.to_dict()
+            logger.debug("State dict: %s", serialized_state)
+            context.set_state(serialized_state)
             logger.info(f"[entity_function] Operation {operation} completed successfully")
 
         except Exception as exc:

python/packages/azurefunctions/agent_framework_azurefunctions/_models.py

@@ -362,50 +362,3 @@ def from_dict(cls, data: dict[str, Any]) -> RunRequest:
             correlation_id=data.get("correlationId"),
             created_at=data.get("created_at"),
         )
-
-
-@dataclass
-class AgentResponse:
-    """Response from agent execution.
-
-    Attributes:
-        response: The agent's text response (or None for structured responses)
-        message: The original message sent to the agent
-        thread_id: The thread identifier
-        status: Status of the execution (success, error, etc.)
-        message_count: Number of messages in the conversation
-        error: Error message if status is error
-        error_type: Type of error if status is error
-        structured_response: Structured response if response_format was provided
-    """
-
-    response: str | None
-    message: str
-    thread_id: str | None
-    status: str
-    message_count: int = 0
-    error: str | None = None
-    error_type: str | None = None
-    structured_response: dict[str, Any] | None = None
-
-    def to_dict(self) -> dict[str, Any]:
-        """Convert to dictionary for JSON serialization."""
-        result: dict[str, Any] = {
-            "message": self.message,
-            "thread_id": self.thread_id,
-            "status": self.status,
-            "message_count": self.message_count,
-        }
-
-        # Add response or structured_response based on what's available
-        if self.structured_response is not None:
-            result["structured_response"] = self.structured_response
-        elif self.response is not None:
-            result["response"] = self.response
-
-        if self.error:
-            result["error"] = self.error
-        if self.error_type:
-            result["error_type"] = self.error_type
-
-        return result

python/packages/azurefunctions/agent_framework_azurefunctions/_orchestration.py

@@ -6,19 +6,27 @@
 """
 
 import uuid
-from collections.abc import AsyncIterator
+from collections.abc import AsyncIterator, Generator
 from typing import TYPE_CHECKING, Any, TypeAlias, cast
 
-from agent_framework import AgentProtocol, AgentRunResponseUpdate, AgentThread, ChatMessage, get_logger
+from agent_framework import (
+    AgentProtocol,
+    AgentRunResponse,
+    AgentRunResponseUpdate,
+    AgentThread,
+    ChatMessage,
+    get_logger,
+)
+from pydantic import BaseModel
 
 from ._models import AgentSessionId, DurableAgentThread, RunRequest
 
 logger = get_logger("agent_framework.azurefunctions.orchestration")
 
 if TYPE_CHECKING:
-    from azure.durable_functions import DurableOrchestrationContext as _DurableOrchestrationContext
+    from azure.durable_functions import DurableOrchestrationContext
 
-    AgentOrchestrationContextType: TypeAlias = _DurableOrchestrationContext
+    AgentOrchestrationContextType: TypeAlias = DurableOrchestrationContext
 else:
     AgentOrchestrationContextType = Any
 
@@ -81,13 +89,18 @@ def description(self) -> str | None:
         """Get the description of the agent."""
         return self._description
 
-    def run(
+    # We return a Generator[Any, Any, ...] here because Durable Functions orchestrations
+    # require yielding Tasks, and the run method must return a Task that can be yielded.
+    # This is an intentional deviation from AgentProtocol which defines run() as async.
+    # The Generator type is correct for Durable Functions orchestrations.
+    def run(  # type: ignore[override]
         self,
         messages: str | ChatMessage | list[str] | list[ChatMessage] | None = None,
         *,
         thread: AgentThread | None = None,
+        response_format: type[BaseModel] | None = None,
         **kwargs: Any,
-    ) -> Any:  # TODO(msft-team): Add a wrapper to respond correctly with `AgentRunResponse`
+    ) -> Generator[Any, Any, AgentRunResponse]:
         """Execute the agent with messages and return a Task for orchestrations.
 
         This method implements AgentProtocol and returns a Task that can be yielded
@@ -96,23 +109,24 @@ def run(
         Args:
             messages: The message(s) to send to the agent
             thread: Optional agent thread for conversation context
-            **kwargs: Additional arguments (enable_tool_calls, response_format, etc.)
+            response_format: Optional Pydantic model for response parsing
+            **kwargs: Additional arguments (enable_tool_calls)
 
         Returns:
-            Task that will resolve to the agent response
+            Yield a task that will resolve to the agent response
 
         Example:
             @app.orchestration_trigger(context_name="context")
             def my_orchestration(context):
                 agent = app.get_agent(context, "MyAgent")
                 thread = agent.get_new_thread()
-                result = yield agent.run("Hello", thread=thread)
+                result = yield from agent.run("Hello", thread=thread)
         """
         message_str = self._normalize_messages(messages)
+        logger.debug(f"[DurableAIAgent] Running agent '{self.agent_name}' with message: {message_str[:100]}...")
 
         # Extract optional parameters from kwargs
         enable_tool_calls = kwargs.get("enable_tool_calls", True)
-        response_format = kwargs.get("response_format")
 
         # Get the session ID for the entity
         if isinstance(thread, DurableAgentThread) and thread.session_id is not None:
@@ -130,6 +144,12 @@ def my_orchestration(context):
         # Generate a deterministic correlation ID for this call
         # This is required by the entity and must be unique per call
         correlation_id = str(self.context.new_uuid())
+        logger.debug(
+            "[DurableAIAgent] Using correlation_id: %s for entity_id: %s for session_id: %s",
+            correlation_id,
+            entity_id,
+            session_id,
+        )
 
         # Prepare the request using RunRequest model
         run_request = RunRequest(
@@ -144,7 +164,57 @@ def my_orchestration(context):
 
         # Call the entity and return the Task directly
         # The orchestration will yield this Task
-        return self.context.call_entity(entity_id, "run_agent", run_request.to_dict())
+        result = yield self.context.call_entity(entity_id, "run_agent", run_request.to_dict())
+
+        logger.debug(
+            "[DurableAIAgent] Entity call completed for correlation_id %s",
+            correlation_id,
+        )
+
+        response = self._load_agent_response(result)
+
+        if response_format is not None:
+            self._ensure_response_format(response_format, correlation_id, response)
+
+        return response
+
+    def _load_agent_response(self, agent_response: AgentRunResponse | dict[str, Any] | None) -> AgentRunResponse:
+        """Convert raw payloads into AgentRunResponse instance."""
+        if agent_response is None:
+            raise ValueError("agent_response cannot be None")
+
+        logger.debug(f"[load_agent_response] Loading agent response of type: {type(agent_response)}")
+
+        if isinstance(agent_response, AgentRunResponse):
+            return agent_response
+        if isinstance(agent_response, dict):
+            logger.debug("[load_agent_response] Converting dict payload using AgentRunResponse.from_dict")
+            return AgentRunResponse.from_dict(agent_response)
+
+        raise TypeError(f"Unsupported type for agent_response: {type(agent_response)}")
+
+    def _ensure_response_format(
+        self,
+        response_format: type[BaseModel] | None,
+        correlation_id: str,
+        response: AgentRunResponse,
+    ) -> None:
+        """Ensure the AgentRunResponse value is parsed into the expected response_format."""
+        if response_format is not None and not isinstance(response.value, response_format):
+            logger.debug(
+                "[DurableAIAgent] Response value type %s does not match expected %s for correlation_id %s",
+                type(response.value),
+                response_format,
+                correlation_id,
+            )
+
+            response.try_parse_value(response_format)
+
+            logger.debug(
+                "[DurableAIAgent] Loaded AgentRunResponse.value for correlation_id %s with type: %s",
+                correlation_id,
+                type(response.value),
+            )
 
     def run_stream(
         self,