chore: fix sampled field semantics on JudgeResult (#137)

Author: jsonbailey

packages/sdk/server-ai/src/ldai/judge/__init__.py

@@ -51,7 +51,7 @@ async def evaluate(
         :param input_text: The input prompt or question that was provided to the AI
         :param output_text: The AI-generated response to be evaluated
         :param sampling_rate: Sampling rate (0-1) to determine if evaluation should be processed (defaults to 1)
-        :return: Evaluation result; ``sampled=True`` when skipped due to sampling rate
+        :return: The result of the judge evaluation.
         """
         judge_result = JudgeResult(judge_config_key=self._ai_config.key)
 
@@ -70,9 +70,9 @@ async def evaluate(
 
             if random.random() > sampling_rate:
                 log.debug(f'Judge evaluation skipped due to sampling rate: {sampling_rate}')
-                judge_result.sampled = True
                 return judge_result
 
+            judge_result.sampled = True
             messages = self._construct_evaluation_messages(input_text, output_text)
             assert self._evaluation_response_structure is not None
 
@@ -110,7 +110,7 @@ async def evaluate_messages(
         :param messages: Array of messages representing the conversation history
         :param response: The AI response to be evaluated
         :param sampling_ratio: Sampling ratio (0-1) to determine if evaluation should be processed (defaults to 1)
-        :return: Evaluation result; ``sampled=True`` when skipped due to sampling rate
+        :return: The result of the judge evaluation.
         """
         input_text = '\r\n'.join([msg.content for msg in messages]) if messages else ''
         output_text = response.message.content

packages/sdk/server-ai/src/ldai/providers/types.py

@@ -65,10 +65,10 @@ class JudgeResult:
     judge_config_key: Optional[str] = None
     success: bool = False
     error_message: Optional[str] = None
-    sampled: bool = False  # True when the judge was skipped due to sampling rate
+    sampled: bool = False  # True when the evaluation was sampled and run
+    metric_key: Optional[str] = None
     score: Optional[float] = None
     reasoning: Optional[str] = None
-    metric_key: Optional[str] = None
 
     def to_dict(self) -> Dict[str, Any]:
         """

packages/sdk/server-ai/src/ldai/tracker.py

@@ -241,6 +241,9 @@ def track_judge_result(self, judge_result: Any) -> None:
 
         :param judge_result: JudgeResult object containing score, metric key, and success status
         """
+        if not judge_result.sampled:
+            return
+
         if judge_result.success and judge_result.metric_key:
             track_data = self.__get_track_data()
             if judge_result.judge_config_key:

packages/sdk/server-ai/tests/test_judge.py

@@ -168,6 +168,7 @@ async def test_evaluate_success_with_valid_response(
 
         assert isinstance(result, JudgeResult)
         assert result.success is True
+        assert result.sampled is True
         assert result.metric_key == '$ld:ai:judge:relevance'
         assert result.score == 0.85
         assert result.reasoning is not None
@@ -194,6 +195,7 @@ async def test_evaluate_success_with_evaluation_response_shape(
 
         assert isinstance(result, JudgeResult)
         assert result.success is True
+        assert result.sampled is True
         assert result.metric_key == '$ld:ai:judge:relevance'
         assert result.score == 0.9
         assert result.reasoning is not None
@@ -288,13 +290,13 @@ async def test_evaluate_handles_exception(
     async def test_evaluate_respects_sampling_rate(
         self, judge_config_with_key: AIJudgeConfig, tracker: LDAIConfigTracker, mock_runner
     ):
-        """Evaluate should return sampled=True when skipped due to sampling rate."""
+        """Evaluate should return sampled=False when skipped due to sampling rate."""
         judge = Judge(judge_config_with_key, tracker, mock_runner)
 
         result = await judge.evaluate("input", "output", sampling_rate=0.0)
 
         assert isinstance(result, JudgeResult)
-        assert result.sampled is True
+        assert result.sampled is False
         assert result.success is False
         mock_runner.invoke_structured_model.assert_not_called()