opik-optimizer 2.0.1__py3-none-any.whl → 2.1.1__py3-none-any.whl

opik_optimizer/__init__.py

@@ -16,8 +16,15 @@ from .gepa_optimizer import GepaOptimizer
 from .logging_config import setup_logging
 from .meta_prompt_optimizer import MetaPromptOptimizer
 from .mipro_optimizer import MiproOptimizer
+from .hierarchical_reflective_optimizer import HierarchicalReflectiveOptimizer
 from .optimization_config.configs import TaskConfig
 from .optimization_result import OptimizationResult
+from .parameter_optimizer import (
+    ParameterOptimizer,
+    ParameterSearchSpace,
+    ParameterSpec,
+    ParameterType,
+)
 
 __version__ = importlib.metadata.version("opik_optimizer")
 
@@ -34,9 +41,14 @@ __all__ = [
     "MetaPromptOptimizer",
     "MiproOptimizer",
     "EvolutionaryOptimizer",
+    "HierarchicalReflectiveOptimizer",
+    "ParameterOptimizer",
     "OptimizationResult",
     "OptimizableAgent",
     "setup_logging",
     "datasets",
     "TaskConfig",
+    "ParameterSearchSpace",
+    "ParameterSpec",
+    "ParameterType",
 ]

opik_optimizer/base_optimizer.py

@@ -470,6 +470,39 @@ class BaseOptimizer(ABC):
             f"{self.__class__.__name__} does not implement optimize_mcp yet."
         )
 
+    def optimize_parameter(
+        self,
+        prompt: "chat_prompt.ChatPrompt",
+        dataset: Dataset,
+        metric: Callable,
+        parameter_space: Any,
+        experiment_config: dict | None = None,
+        n_trials: int | None = None,
+        n_samples: int | None = None,
+        agent_class: type[OptimizableAgent] | None = None,
+        **kwargs: Any,
+    ) -> optimization_result.OptimizationResult:
+        """
+        Optimize LLM call parameters such as temperature or top_k.
+
+        Args:
+            prompt: The chat prompt to evaluate with tuned parameters
+            dataset: Dataset providing evaluation examples
+            metric: Objective function to maximize
+            parameter_space: Definition of the search space for tunable parameters
+            experiment_config: Optional experiment metadata
+            n_trials: Number of trials to run (optimizer specific default if None)
+            n_samples: Number of dataset samples to evaluate per trial (None for all)
+            agent_class: Optional custom agent class to execute evaluations
+            **kwargs: Additional optimizer specific settings
+
+        Returns:
+            OptimizationResult: Structured result describing the best parameters found
+        """
+        raise NotImplementedError(
+            f"{self.__class__.__name__} does not implement optimize_parameter yet."
+        )
+
     def get_history(self) -> list[OptimizationRound]:
         """
         Get the optimization history.

opik_optimizer/hierarchical_reflective_optimizer/__init__.py

@@ -0,0 +1,5 @@
+from .hierarchical_reflective_optimizer import HierarchicalReflectiveOptimizer
+
+__all__ = [
+    "HierarchicalReflectiveOptimizer",
+]