fix(items): strip created_by when replaying RunItems as input

src/agents/items.py

@@ -150,14 +150,7 @@ def _get_agent_via_weakref(self, attr_name: str, ref_name: str) -> Any:
 
     def to_input_item(self) -> TResponseInputItem:
         """Converts this item into an input item suitable for passing to the model."""
-        if isinstance(self.raw_item, dict):
-            # We know that input items are dicts, so we can ignore the type error
-            return self.raw_item  # type: ignore
-        elif isinstance(self.raw_item, BaseModel):
-            # All output items are Pydantic models that can be converted to input items.
-            return self.raw_item.model_dump(exclude_unset=True)  # type: ignore
-        else:
-            raise AgentsException(f"Unexpected raw item type: {type(self.raw_item)}")
+        return _output_item_to_input_item(self.raw_item)
 
 
 @dataclass
@@ -491,6 +484,7 @@ def to_input_item(self) -> TResponseInputItem:
                         if isinstance(outcome, dict):
                             if outcome.get("type") == "exit":
                                 entry["outcome"] = outcome
+            payload.pop("created_by", None)
             return cast(TResponseInputItem, payload)
 
         return super().to_input_item()

tests/test_items_helpers.py

@@ -621,6 +621,46 @@ def test_to_input_items_strips_created_by_for_non_tool_search_items() -> None:
     assert all("created_by" not in item for item in input_items)
 
 
+def test_tool_call_item_to_input_item_strips_created_by() -> None:
+    """RunItem replay must strip output-only created_by, matching ModelResponse.to_input_items."""
+    agent = Agent(name="A")
+    call = ResponseFunctionToolCall.model_validate(
+        {
+            "id": "fc_1",
+            "arguments": "{}",
+            "call_id": "call_1",
+            "name": "lookup",
+            "type": "function_call",
+            "created_by": "server",
+        }
+    )
+    item = ToolCallItem(agent=agent, raw_item=call)
+    replayed = item.to_input_item()
+    assert isinstance(replayed, dict)
+    assert replayed["type"] == "function_call"
+    assert "created_by" not in replayed
+    assert getattr(call, "created_by", None) == "server"
+
+
+def test_tool_call_output_item_to_input_item_strips_created_by() -> None:
+    agent = Agent(name="A")
+    item = ToolCallOutputItem(
+        agent=agent,
+        raw_item={
+            "type": "function_call_output",
+            "call_id": "call_1",
+            "output": "ok",
+            "created_by": "server",
+        },
+        output="ok",
+    )
+    replayed = item.to_input_item()
+    assert isinstance(replayed, dict)
+    assert replayed["type"] == "function_call_output"
+    assert "created_by" not in replayed
+    assert item.raw_item["created_by"] == "server"
+
+
 def test_to_input_items_strips_nested_created_by_from_shell_call_output() -> None:
     """``shell_call_output`` carries ``created_by`` at the item level and inside each output chunk.
 

tests/test_run_internal_items.py

@@ -898,6 +898,29 @@ def test_run_item_to_input_item_strips_tool_search_created_by() -> None:
     assert "created_by" not in converted_output
 
 
+def test_run_item_to_input_item_strips_function_call_created_by() -> None:
+    agent = Agent(name="A")
+    tool_call = ToolCallItem(
+        agent=agent,
+        raw_item=ResponseFunctionToolCall.model_validate(
+            {
+                "id": "fc_1",
+                "arguments": "{}",
+                "call_id": "call_1",
+                "name": "lookup",
+                "type": "function_call",
+                "created_by": "server",
+            }
+        ),
+    )
+
+    converted = run_items.run_item_to_input_item(tool_call)
+
+    assert isinstance(converted, dict)
+    assert converted["type"] == "function_call"
+    assert "created_by" not in converted
+
+
 def test_run_item_to_input_item_omits_tool_call_metadata() -> None:
     agent = Agent(name="A")
     tool_call = ToolCallItem(