feat: add kwargs in evaluate function of agent and common grader (#159)

Author: jc200808

openjudge/graders/agent/action/action_alignment.py

@@ -215,6 +215,7 @@ async def _aevaluate(
         action: str,
         history: Optional[List[Dict[str, Any]]] = None,
         context: Optional[str] = None,
+        **kwargs: Any,
     ) -> GraderScore:
         """
         Evaluate action alignment with plan

openjudge/graders/agent/action/action_loop.py

@@ -50,6 +50,7 @@ def __init__(
     async def _aevaluate(
         self,
         messages: List[Dict[str, Any]],
+        **kwargs: Any,
     ) -> GraderScore:
         """
         Detect loops in action sequences by comparing all pairs of action signatures.

openjudge/graders/agent/observation/observation_information_gain.py

@@ -60,6 +60,7 @@ def __init__(
     async def _aevaluate(
         self,
         messages: List[Dict[str, Any]],
+        **kwargs: Any,
     ) -> GraderScore:
         """
         Evaluate information gain and redundancy in observation observations.

openjudge/graders/agent/tool/tool_call_accuracy.py

@@ -280,6 +280,7 @@ async def _aevaluate(
         tool_definitions: Dict[str, Any] | List[Dict[str, Any]],
         tool_calls: Dict[str, Any] | List[Dict[str, Any]] | None = None,
         response: str | List[Dict[str, Any]] | None = None,
+        **kwargs: Any,
     ) -> GraderScore | GraderError:
         """
         Evaluate tool call accuracy

openjudge/graders/agent/tool/tool_call_precision_recall_match.py

@@ -199,6 +199,7 @@ async def _aevaluate(
         self,
         tool_calls: List[Dict[str, Any]],
         reference_tool_calls: List[Dict[str, Any]],
+        **kwargs: Any,
     ) -> GraderScore | GraderError:
         """
         Evaluate tool call precision/recall against reference.

openjudge/graders/agent/tool/tool_call_step_sequence_match.py

@@ -358,6 +358,7 @@ async def _aevaluate(
         self,
         messages: List[Dict[str, Any]],
         reference_tool_calls: List[List[Dict[str, Any]]],
+        **kwargs: Any,
     ) -> GraderScore | GraderError:
         """
         Evaluate tool call sequence matching against reference.

openjudge/graders/agent/tool/tool_call_success.py

@@ -267,6 +267,7 @@ async def _aevaluate(
         tool_definitions: Union[Dict[str, Any], List[Dict[str, Any]]],
         tool_calls: Union[Dict[str, Any], List[Dict[str, Any]]],
         tool_responses: Union[str, List[str]],
+        **kwargs: Any,
     ) -> GraderScore:
         """
         Evaluate tool call success

openjudge/graders/agent/tool/tool_parameter_check.py

@@ -214,6 +214,7 @@ async def _aevaluate(
         query: Union[str, List[Dict[str, Any]]],
         tool_definitions: Union[Dict[str, Any], List[Dict[str, Any]]],
         tool_calls: Union[Dict[str, Any], List[Dict[str, Any]]],
+        **kwargs: Any,
     ) -> GraderScore:
         """
         Evaluate tool parameter extraction correctness

openjudge/graders/agent/tool/tool_selection.py

@@ -230,6 +230,7 @@ async def _aevaluate(
         query: Union[str, List[Dict[str, Any]]],
         tool_definitions: Union[Dict[str, Any], List[Dict[str, Any]]],
         tool_calls: Union[Dict[str, Any], List[Dict[str, Any]]],
+        **kwargs: Any,
     ) -> GraderScore:
         """
         Evaluate tool selection

openjudge/graders/agent/trajectory/trajectory_accuracy.py

@@ -270,6 +270,7 @@ def _format_messages(
     async def _aevaluate(
         self,
         messages: List[Dict[str, Any]],
+        **kwargs: Any,
     ) -> GraderScore | GraderError:
         """
         Evaluate trajectory accuracy