fix: add system_prompt to structured_output_span before adding input_messages (#709)

* fix:  add system_prompt  to structured_output_span  before adding input_messages

* test: Add system message ordering validation to agent structured output test

* Switch to ensuring exact ordering of messages

---------

Co-authored-by: Dennis Tsai (RD-AS) <[email protected]>
Co-authored-by: Mackenzie Zastrow <[email protected]>

Author: 3 people

src/strands/agent/agent.py

@@ -470,16 +470,16 @@ async def structured_output_async(
                         "gen_ai.operation.name": "execute_structured_output",
                     }
                 )
-                for message in temp_messages:
-                    structured_output_span.add_event(
-                        f"gen_ai.{message['role']}.message",
-                        attributes={"role": message["role"], "content": serialize(message["content"])},
-                    )
                 if self.system_prompt:
                     structured_output_span.add_event(
                         "gen_ai.system.message",
                         attributes={"role": "system", "content": serialize([{"text": self.system_prompt}])},
                     )
+                for message in temp_messages:
+                    structured_output_span.add_event(
+                        f"gen_ai.{message['role']}.message",
+                        attributes={"role": message["role"], "content": serialize(message["content"])},
+                    )
                 events = self.model.structured_output(output_model, temp_messages, system_prompt=self.system_prompt)
                 async for event in events:
                     if "callback" in event:

tests/strands/agent/test_agent.py

@@ -18,6 +18,7 @@
 from strands.handlers.callback_handler import PrintingCallbackHandler, null_callback_handler
 from strands.models.bedrock import DEFAULT_BEDROCK_MODEL_ID, BedrockModel
 from strands.session.repository_session_manager import RepositorySessionManager
+from strands.telemetry.tracer import serialize
 from strands.types.content import Messages
 from strands.types.exceptions import ContextWindowOverflowException, EventLoopException
 from strands.types.session import Session, SessionAgent, SessionMessage, SessionType
@@ -1028,15 +1029,23 @@ def test_agent_structured_output(agent, system_prompt, user, agenerator):
         }
     )
 
-    mock_span.add_event.assert_any_call(
-        "gen_ai.user.message",
-        attributes={"role": "user", "content": '[{"text": "Jane Doe is 30 years old and her email is [email protected]"}]'},
-    )
+    # ensure correct otel event messages are emitted
+    act_event_names = mock_span.add_event.call_args_list
+    exp_event_names = [
+        unittest.mock.call(
+            "gen_ai.system.message", attributes={"role": "system", "content": serialize([{"text": system_prompt}])}
+        ),
+        unittest.mock.call(
+            "gen_ai.user.message",
+            attributes={
+                "role": "user",
+                "content": '[{"text": "Jane Doe is 30 years old and her email is [email protected]"}]',
+            },
+        ),
+        unittest.mock.call("gen_ai.choice", attributes={"message": json.dumps(user.model_dump())}),
+    ]
 
-    mock_span.add_event.assert_called_with(
-        "gen_ai.choice",
-        attributes={"message": json.dumps(user.model_dump())},
-    )
+    assert act_event_names == exp_event_names
 
 
 def test_agent_structured_output_multi_modal_input(agent, system_prompt, user, agenerator):