[evaluation] Name both SDK and asset aliases in TaskNavigationEfficiency validator errors (#48048)

* [evaluation] Name both SDK and asset aliases in TaskNavigationEfficiency validator errors

TaskNavigationEfficiencyValidator accepts either the SDK input names (response/ground_truth) or the azureml-assets names (actions/expected_actions), but its user-facing validation messages referenced only response/ground_truth. A caller who correctly supplied actions/expected_actions could get an error naming ground_truth/response - parameters they never used and that the eval service does not accept as data-mapping keys - which was confusing.

Update every validation message that names response/ground_truth to also cite the accepted alias, e.g. 'ground_truth' (also accepted as 'expected_actions') and 'response' (also accepted as 'actions'). Message text only; no changes to categories, logic, or control flow.

* [evaluation] Lead TaskNavigationEfficiency validator errors with azureml-assets names

Swap the alias order in every TaskNavigationEfficiencyValidator error message so the azureml-assets input names ('actions'/'expected_actions') are primary and the SDK names ('response'/'ground_truth') are the parenthetical alias. Update the CHANGELOG entry to match. Message text only; no logic, category, or control-flow changes.

* Add test

Author: m7md7sien

sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluators/_common/_validators/_task_navigation_efficiency_validator.py

@@ -53,15 +53,15 @@ def _validate_response(self, response: Any) -> Optional[EvaluationException]:
         """Validate the response parameter."""
         if response is None:
             return EvaluationException(
-                message="'response' parameter is required and cannot be None.",
+                message="'actions' (also accepted as 'response') parameter is required and cannot be None.",
                 blame=ErrorBlame.USER_ERROR,
                 category=ErrorCategory.MISSING_FIELD,
                 target=self.error_target,
             )
 
         if not isinstance(response, list):
             return EvaluationException(
-                message="'response' must be a list of messages.",
+                message="'actions' (also accepted as 'response') must be a list of messages.",
                 blame=ErrorBlame.USER_ERROR,
                 category=ErrorCategory.INVALID_VALUE,
                 target=self.error_target,
@@ -141,7 +141,7 @@ def _validate_ground_truth(self, ground_truth: Any) -> Optional[EvaluationExcept
         """Validate the ground_truth parameter."""
         if not ground_truth:
             return EvaluationException(
-                message="'ground_truth' parameter is required and cannot be None or empty.",
+                message="'expected_actions' (also accepted as 'ground_truth') parameter is required and cannot be None or empty.",
                 blame=ErrorBlame.USER_ERROR,
                 category=ErrorCategory.INVALID_VALUE,
                 target=self.error_target,
@@ -155,7 +155,7 @@ def _validate_ground_truth(self, ground_truth: Any) -> Optional[EvaluationExcept
             # Validate tuple format: (list, dict)
             if len(ground_truth) != 2:
                 return EvaluationException(
-                    message="When 'ground_truth' is a tuple, it must contain exactly 2 elements: (tool_names_list, parameters_dict).",
+                    message="When 'expected_actions' (also accepted as 'ground_truth') is a tuple, it must contain exactly 2 elements: (tool_names_list, parameters_dict).",
                     blame=ErrorBlame.USER_ERROR,
                     category=ErrorCategory.INVALID_VALUE,
                     target=self.error_target,
@@ -166,15 +166,15 @@ def _validate_ground_truth(self, ground_truth: Any) -> Optional[EvaluationExcept
             # Validate tool names list
             if not isinstance(tool_names, list):
                 return EvaluationException(
-                    message="First element of 'ground_truth' tuple must be a list of tool names.",
+                    message="First element of 'expected_actions' (also accepted as 'ground_truth') tuple must be a list of tool names.",
                     blame=ErrorBlame.USER_ERROR,
                     category=ErrorCategory.INVALID_VALUE,
                     target=self.error_target,
                 )
 
             if len(tool_names) == 0:
                 return EvaluationException(
-                    message="Tool names list in 'ground_truth' cannot be empty.",
+                    message="Tool names list in 'expected_actions' (also accepted as 'ground_truth') cannot be empty.",
                     blame=ErrorBlame.USER_ERROR,
                     category=ErrorCategory.INVALID_VALUE,
                     target=self.error_target,
@@ -183,7 +183,7 @@ def _validate_ground_truth(self, ground_truth: Any) -> Optional[EvaluationExcept
             for idx, name in enumerate(tool_names):
                 if not isinstance(name, str):
                     return EvaluationException(
-                        message=f"Tool name at index {idx} in 'ground_truth' must be a string, got {type(name).__name__}.",
+                        message=f"Tool name at index {idx} in 'expected_actions' (also accepted as 'ground_truth') must be a string, got {type(name).__name__}.",
                         blame=ErrorBlame.USER_ERROR,
                         category=ErrorCategory.INVALID_VALUE,
                         target=self.error_target,
@@ -192,7 +192,7 @@ def _validate_ground_truth(self, ground_truth: Any) -> Optional[EvaluationExcept
             # Validate parameters dict
             if not isinstance(parameters, dict):
                 return EvaluationException(
-                    message="Second element of 'ground_truth' tuple must be a dictionary of parameters.",
+                    message="Second element of 'expected_actions' (also accepted as 'ground_truth') tuple must be a dictionary of parameters.",
                     blame=ErrorBlame.USER_ERROR,
                     category=ErrorCategory.INVALID_VALUE,
                     target=self.error_target,
@@ -202,7 +202,7 @@ def _validate_ground_truth(self, ground_truth: Any) -> Optional[EvaluationExcept
             for tool_name, params in parameters.items():
                 if not isinstance(params, dict):
                     return EvaluationException(
-                        message=f"Parameters for tool '{tool_name}' in 'ground_truth' must be a dictionary, got {type(params).__name__}.",
+                        message=f"Parameters for tool '{tool_name}' in 'expected_actions' (also accepted as 'ground_truth') must be a dictionary, got {type(params).__name__}.",
                         blame=ErrorBlame.USER_ERROR,
                         category=ErrorCategory.INVALID_VALUE,
                         target=self.error_target,
@@ -212,7 +212,7 @@ def _validate_ground_truth(self, ground_truth: Any) -> Optional[EvaluationExcept
             # Validate list of tool names
             if len(ground_truth) == 0:
                 return EvaluationException(
-                    message="'ground_truth' list cannot be empty.",
+                    message="'expected_actions' (also accepted as 'ground_truth') list cannot be empty.",
                     blame=ErrorBlame.USER_ERROR,
                     category=ErrorCategory.INVALID_VALUE,
                     target=self.error_target,
@@ -229,7 +229,7 @@ def _validate_ground_truth(self, ground_truth: Any) -> Optional[EvaluationExcept
 
         else:
             return EvaluationException(
-                message="'ground_truth' must be either a list of tool names or a tuple of (tool_names_list, parameters_dict).",
+                message="'expected_actions' (also accepted as 'ground_truth') must be either a list of tool names or a tuple of (tool_names_list, parameters_dict).",
                 blame=ErrorBlame.USER_ERROR,
                 category=ErrorCategory.INVALID_VALUE,
                 target=self.error_target,

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

@@ -513,6 +513,29 @@ def test_ground_truth_tuple_params_not_dict_raises(self):
         with pytest.raises(EvaluationException):
             validator.validate_eval_input(eval_input)
 
+    @pytest.mark.parametrize(
+        "eval_input, canonical_name, alias_name",
+        [
+            # 'response'/'actions' missing (None).
+            ({"response": None, "ground_truth": ["search"]}, "response", "actions"),
+            # 'response'/'actions' wrong type.
+            ({"response": "not a list", "ground_truth": ["search"]}, "response", "actions"),
+            # 'ground_truth'/'expected_actions' empty -- the reported UX case.
+            ({"response": [{"role": "user", "content": "x"}], "ground_truth": []}, "ground_truth", "expected_actions"),
+            # 'ground_truth'/'expected_actions' wrong type.
+            ({"response": [{"role": "user", "content": "x"}], "ground_truth": 123}, "ground_truth", "expected_actions"),
+        ],
+    )
+    def test_error_messages_name_both_canonical_and_alias(self, eval_input, canonical_name, alias_name):
+        # Regression guard: user-facing validation errors must cite both the SDK input name and
+        # its azureml-assets alias, so a caller never sees a parameter name they did not supply.
+        validator = TaskNavigationEfficiencyValidator(error_target=TARGET)
+        with pytest.raises(EvaluationException) as exc_info:
+            validator.validate_eval_input(eval_input)
+        message = str(exc_info.value)
+        assert f"'{canonical_name}'" in message
+        assert f"'{alias_name}'" in message
+
 
 @pytest.mark.unittest
 class TestMessagesOrQueryResponseInputValidator: