feat!: remove evaluation_metric_keys plural field from AIJudgeConfig and AIJudgeConfigDefault

The plural `evaluation_metric_keys` field is removed from both
`AIJudgeConfig` and `AIJudgeConfigDefault`. Defaults should only set
the singular `evaluation_metric_key`. The client read path still
accepts the plural `evaluationMetricKeys` from a variation payload and
falls back to its first element when the singular is not set, so older
flag values continue to resolve a metric key.

BREAKING CHANGE: `evaluation_metric_keys` is no longer a field on
`AIJudgeConfig` or `AIJudgeConfigDefault`. Construct these dataclasses
with `evaluation_metric_key` (singular) instead. The wire-format
`evaluationMetricKeys` plural is still read from variations for
backward compatibility with older AI configs and is mapped onto the
singular field on the returned `AIJudgeConfig`.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: jsonbailey

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

@@ -334,8 +334,6 @@ class AIJudgeConfigDefault(AIConfigDefault):
     Default Judge-specific AI Config with required evaluation metric key.
     """
     messages: Optional[List[LDMessage]] = None
-    # Deprecated: evaluation_metric_key is used instead
-    evaluation_metric_keys: Optional[List[str]] = None
     evaluation_metric_key: Optional[str] = None
 
     def to_dict(self) -> dict:
@@ -345,9 +343,6 @@ def to_dict(self) -> dict:
         result = self._base_to_dict()
         result['messages'] = [message.to_dict() for message in self.messages] if self.messages else None
         result['evaluationMetricKey'] = self.evaluation_metric_key
-        # Include deprecated evaluationMetricKeys for backward compatibility
-        if self.evaluation_metric_keys:
-            result['evaluationMetricKeys'] = self.evaluation_metric_keys
         return result
 
 
@@ -356,8 +351,6 @@ class AIJudgeConfig(AIConfig):
     """
     Judge-specific AI Config with required evaluation metric key.
     """
-    # Deprecated: evaluation_metric_key is used instead
-    evaluation_metric_keys: List[str] = field(default_factory=list)
     messages: Optional[List[LDMessage]] = None
     evaluation_metric_key: Optional[str] = None