fix(litellm): enhance structured output handling

src/strands/models/litellm.py

@@ -238,8 +238,8 @@ async def _structured_output_using_response_schema(
 
         if len(response.choices) > 1:
             raise ValueError("Multiple choices found in the response.")
-        if not response.choices or response.choices[0].finish_reason != "tool_calls":
-            raise ValueError("No tool_calls found in response")
+        if not response.choices:
+            raise ValueError("No choices found in response")
 
         choice = response.choices[0]
         try:

tests/strands/models/test_litellm.py

@@ -277,6 +277,13 @@ async def test_structured_output(litellm_acompletion, model, test_output_model_c
     mock_choice = unittest.mock.Mock()
     mock_choice.finish_reason = "tool_calls"
     mock_choice.message.content = '{"name": "John", "age": 30}'
+    # PATCH START: mock tool_calls as list with .function.arguments
+    tool_call_mock = unittest.mock.Mock()
+    tool_call_function_mock = unittest.mock.Mock()
+    tool_call_function_mock.arguments = '{"name": "John", "age": 30}'
+    tool_call_mock.function = tool_call_function_mock
+    mock_choice.message.tool_calls = [tool_call_mock]
+    # PATCH END
     mock_response = unittest.mock.Mock()
     mock_response.choices = [mock_choice]