(engsys) ci,fix: Fix Run Black falsely reporting success (Azure#38095)

* ci,fix(run_black): Resolve package directory when service dir is "auto"

* style: Run black

Author: kdestin

scripts/devops_tasks/validate_formatting.py

@@ -23,9 +23,12 @@ def run_black(glob_string, service_dir):
     results = []
     logging.info("Running black for {}".format(service_dir))
 
-    discovered_packages = discover_targeted_packages(
-        glob_string, os.path.join(root_dir, "sdk", service_dir)
-    )
+    if service_dir and service_dir != "auto":
+        target_dir = os.path.join(root_dir, "sdk", service_dir)
+    else:
+        target_dir = root_dir
+
+    discovered_packages = discover_targeted_packages(glob_string, target_dir)
 
     for package in discovered_packages:
         package_name = os.path.basename(package)
@@ -43,7 +46,7 @@ def run_black(glob_string, service_dir):
                     "-e",
                     "black",
                     "--",
-                    os.path.join("sdk", service_dir, package_name),
+                    package,
                 ],
                 stdout=subprocess.PIPE,
                 stderr=subprocess.STDOUT,

sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/simulator/_indirect_attack_simulator.py

@@ -191,7 +191,7 @@ async def __call__(
             template_parameters = completed_task.get("template_parameters", {})  # type: ignore
             xpia_attack_type = template_parameters.get("xpia_attack_type", "")  # type: ignore
             action = template_parameters.get("action", "")  # type: ignore
-            document_type = template_parameters.get("document_type", "")    # type: ignore
+            document_type = template_parameters.get("document_type", "")  # type: ignore
             sim_results.append(
                 {
                     "messages": completed_task["messages"],  # type: ignore

sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/simulator/_simulator.py

@@ -236,15 +236,19 @@ async def run_simulation(simulation: List[Union[str, Dict[str, Any]]]) -> JsonLi
                         user_turn = Turn(
                             role=ConversationRole.USER,
                             content=str(simulated_turn.get("content")),
-                            context=str(simulated_turn.get("context"))
+                            context=str(simulated_turn.get("context")),
                         )
                     else:
-                        raise ValueError("Each simulated turn must be a string or a dict with 'content' and 'context' keys")
+                        raise ValueError(
+                            "Each simulated turn must be a string or a dict with 'content' and 'context' keys"
+                        )
                     current_simulation.add_to_history(user_turn)
                     assistant_response, assistant_context = await self._get_target_response(
                         target=target, api_call_delay_sec=api_call_delay_sec, conversation_history=current_simulation
                     )
-                    assistant_turn = Turn(role=ConversationRole.ASSISTANT, content=assistant_response, context=assistant_context)
+                    assistant_turn = Turn(
+                        role=ConversationRole.ASSISTANT, content=assistant_response, context=assistant_context
+                    )
                     current_simulation.add_to_history(assistant_turn)
                     async with progress_bar_lock:
                         progress_bar.update(1)
@@ -286,7 +290,7 @@ async def _extend_conversation_with_simulator(
         prompty_model_config: Dict[str, Any],
         target: Callable,
         progress_bar: tqdm,
-        progress_bar_lock: asyncio.Lock
+        progress_bar_lock: asyncio.Lock,
     ):
         """
         Extends an ongoing conversation using a user simulator until the maximum number of turns is reached.
@@ -329,7 +333,9 @@ async def _extend_conversation_with_simulator(
             assistant_response, assistant_context = await self._get_target_response(
                 target=target, api_call_delay_sec=api_call_delay_sec, conversation_history=current_simulation
             )
-            assistant_turn = Turn(role=ConversationRole.ASSISTANT, content=assistant_response, context=assistant_context)
+            assistant_turn = Turn(
+                role=ConversationRole.ASSISTANT, content=assistant_response, context=assistant_context
+            )
             current_simulation.add_to_history(assistant_turn)
             async with progress_bar_lock:
                 progress_bar.update(1)
@@ -667,7 +673,9 @@ async def _complete_conversation(
             assistant_response, assistant_context = await self._get_target_response(
                 target=target, api_call_delay_sec=api_call_delay_sec, conversation_history=conversation_history
             )
-            assistant_turn = Turn(role=ConversationRole.ASSISTANT, content=assistant_response, context=assistant_context)
+            assistant_turn = Turn(
+                role=ConversationRole.ASSISTANT, content=assistant_response, context=assistant_context
+            )
             conversation_history.add_to_history(assistant_turn)
             progress_bar.update(1)
 

sdk/evaluation/azure-ai-evaluation/tests/unittests/test_built_in_evaluator.py

@@ -59,14 +59,17 @@ def test_retrieval_evaluator_keys(self, mock_model_config):
         conversation = {
             "messages": [
                 {"role": "user", "content": "What is the value of 2 + 2?"},
-                {"role": "assistant", "content": "2 + 2 = 4", "context": {
-                    "citations": [
+                {
+                    "role": "assistant",
+                    "content": "2 + 2 = 4",
+                    "context": {
+                        "citations": [
                             {"id": "math_doc.md", "content": "Information about additions: 1 + 2 = 3, 2 + 2 = 4"}
                         ]
-                    }
-                }
+                    },
+                },
             ]
         }
 
         result = retrieval_eval(conversation=conversation)
-        assert result["retrieval"] == result["gpt_retrieval"] == 1
+        assert result["retrieval"] == result["gpt_retrieval"] == 1