fix tests

llama_stack/providers/inline/agents/meta_reference/responses/openai_responses.py

@@ -345,7 +345,7 @@ class OpenAIResponsesImpl:
                 return
 
         # Structured outputs
-        response_format = convert_response_text_to_chat_response_format(text)
+        response_format = await convert_response_text_to_chat_response_format(text)
 
         ctx = ChatCompletionContext(
             model=model,

llama_stack/providers/inline/agents/meta_reference/responses/streaming.py

@@ -13,8 +13,8 @@ from llama_stack.apis.agents.openai_responses import (
     ApprovalFilter,
     MCPListToolsTool,
     OpenAIResponseContentPartOutputText,
-    OpenAIResponseError,
     OpenAIResponseContentPartRefusal,
+    OpenAIResponseError,
     OpenAIResponseInputTool,
     OpenAIResponseInputToolMCP,
     OpenAIResponseMCPApprovalRequest,
@@ -226,6 +226,11 @@ class StreamingResponseOrchestrator:
                         completion_result_data = stream_event_or_result
                     else:
                         yield stream_event_or_result
+
+                # If violation detected, skip the rest of processing since we already sent refusal
+                if self.violation_detected:
+                    return
+
                 if not completion_result_data:
                     raise ValueError("Streaming chunk processor failed to return completion data")
                 last_completion_result = completion_result_data

llama_stack/providers/inline/agents/meta_reference/responses/utils.py

@@ -178,7 +178,7 @@ async def convert_response_input_to_chat_messages(
                 pass
             else:
                 content = await convert_response_content_to_chat_content(input_item.content)
-                message_type = get_message_type_by_role(input_item.role)
+                message_type = await get_message_type_by_role(input_item.role)
                 if message_type is None:
                     raise ValueError(
                         f"Llama Stack OpenAI Responses does not yet support message role '{input_item.role}' in this context"

tests/unit/providers/agents/meta_reference/test_responses_safety_utils.py

@@ -84,14 +84,14 @@ def test_extract_shield_ids_empty_list(responses_impl):
     assert result == []
 
 
-def test_extract_shield_ids_unknown_format(responses_impl, caplog):
-    """Test extraction with unknown shield format logs warning."""
+def test_extract_shield_ids_unknown_format(responses_impl):
+    """Test extraction with unknown shield format raises ValueError."""
     # Create an object that's neither string nor ResponseShieldSpec
     unknown_object = {"invalid": "format"}  # Plain dict, not ResponseShieldSpec
     shields = ["valid-shield", unknown_object, "another-shield"]
-    result = extract_shield_ids(shields)
-    assert result == ["valid-shield", "another-shield"]
-    assert "Unknown shield format" in caplog.text
+
+    with pytest.raises(ValueError, match="Unknown shield format.*expected str or ResponseShieldSpec"):
+        extract_shield_ids(shields)
 
 
 def test_extract_text_content_string(responses_impl):