opik-optimizer 2.1.3__py3-none-any.whl → 2.2.0__py3-none-any.whl

opik_optimizer/evolutionary_optimizer/evaluation_ops.py

@@ -6,6 +6,7 @@ from .. import task_evaluator
 from ..optimization_config import mappers, chat_prompt
 from ..mcp_utils.mcp_workflow import MCPExecutionConfig
 import opik
+from opik import opik_context
 import copy
 
 if TYPE_CHECKING:  # pragma: no cover - typing only
@@ -15,7 +16,7 @@ if TYPE_CHECKING:  # pragma: no cover - typing only
 class EvaluationOps:
     if TYPE_CHECKING:
         agent_class: type[Any]
-        num_threads: int
+        n_threads: int
 
     def _evaluate_prompt(
         self,
@@ -79,6 +80,16 @@ class EvaluationOps:
 
             if mcp_execution_config is None:
                 model_output = agent.invoke(messages)
+
+                # Add tags to trace for optimization tracking
+                if (
+                    hasattr(self, "current_optimization_id")
+                    and self.current_optimization_id
+                ):
+                    opik_context.update_current_trace(
+                        tags=[self.current_optimization_id, "Evaluation"]
+                    )
+
                 return {mappers.EVALUATED_LLM_TASK_OUTPUT: model_output}
 
             coordinator = mcp_execution_config.coordinator
@@ -119,6 +130,15 @@ class EvaluationOps:
             else:
                 final_response = raw_model_output
 
+            # Add tags to trace for optimization tracking
+            if (
+                hasattr(self, "current_optimization_id")
+                and self.current_optimization_id
+            ):
+                opik_context.update_current_trace(
+                    tags=[self.current_optimization_id, "Evaluation"]
+                )
+
             return {mappers.EVALUATED_LLM_TASK_OUTPUT: final_response.strip()}
 
         score = task_evaluator.evaluate(
@@ -126,8 +146,8 @@ class EvaluationOps:
             dataset_item_ids=dataset_item_ids,
             metric=metric,
             evaluated_task=llm_task,
-            num_threads=self.num_threads,
-            project_name=experiment_config.get("project_name"),
+            num_threads=self.n_threads,
+            project_name=optimizer.project_name,
             n_samples=n_samples if dataset_item_ids is None else None,
             experiment_config=experiment_config,
             optimization_id=optimization_id,

opik_optimizer/evolutionary_optimizer/evolutionary_optimizer.py

@@ -31,7 +31,6 @@ from opik_optimizer.utils.prompt_segments import extract_prompt_segments
 from .mcp import EvolutionaryMCPContext, finalize_mcp_result
 
 from . import reporting
-from .llm_support import LlmSupport
 from .mutation_ops import MutationOps
 from .crossover_ops import CrossoverOps
 from .population_ops import PopulationOps
@@ -48,19 +47,41 @@ creator = cast(Any, _creator)  # type: ignore[assignment]
 
 class EvolutionaryOptimizer(BaseOptimizer):
     """
-    The Evolutionary Optimizer can be used to optimize prompts using a 4 stage genetic algorithm
-    approach:
+    Evolutionary Optimizer that uses genetic algorithms to evolve and improve prompts over generations.
 
-    1. Generate a set of candidate prompts based on variations of the best prompts (exploitation) as
-    well as completely new prompts (exploration)
-    2. Evaluate the candidate prompts
-    3. Select the best prompts
-    4. Repeat until convergence
+    This optimizer uses a 4-stage genetic algorithm approach:
+
+    1. Generate candidate prompts through variations of the best prompts (exploitation) and
+       completely new prompts (exploration)
+    2. Evaluate the candidate prompts on the dataset
+    3. Select the best prompts based on fitness
+    4. Repeat until convergence or max generations reached
 
     This algorithm is best used if you have a first draft prompt and would like to find a better
-    prompt.
+    prompt through iterative evolution. It supports both single-objective and multi-objective
+    optimization (balancing performance and prompt length).
 
     Note: This algorithm is time consuming and can be expensive to run.
+
+    Args:
+        model: LiteLLM model name for optimizer's internal operations (mutations, crossover, etc.)
+        model_parameters: Optional dict of LiteLLM parameters for optimizer's internal LLM calls.
+            Common params: temperature, max_tokens, max_completion_tokens, top_p.
+            See: https://docs.litellm.ai/docs/completion/input
+        population_size: Number of prompts in the population
+        num_generations: Number of generations to run
+        mutation_rate: Mutation rate for genetic operations
+        crossover_rate: Crossover rate for genetic operations
+        tournament_size: Tournament size for selection
+        elitism_size: Number of elite prompts to preserve across generations
+        adaptive_mutation: Whether to use adaptive mutation that adjusts based on population diversity
+        enable_moo: Whether to enable multi-objective optimization (optimizes metric and prompt length)
+        enable_llm_crossover: Whether to enable LLM-based crossover operations
+        output_style_guidance: Optional guidance for output style in generated prompts
+        infer_output_style: Whether to automatically infer output style from the dataset
+        n_threads: Number of threads for parallel evaluation
+        verbose: Controls internal logging/progress bars (0=off, 1=on)
+        seed: Random seed for reproducibility
     """
 
     DEFAULT_POPULATION_SIZE = 30
@@ -98,43 +119,23 @@ class EvolutionaryOptimizer(BaseOptimizer):
 
     def __init__(
         self,
-        model: str,
+        model: str = "gpt-4o",
+        model_parameters: dict[str, Any] | None = None,
         population_size: int = DEFAULT_POPULATION_SIZE,
         num_generations: int = DEFAULT_NUM_GENERATIONS,
         mutation_rate: float = DEFAULT_MUTATION_RATE,
         crossover_rate: float = DEFAULT_CROSSOVER_RATE,
         tournament_size: int = DEFAULT_TOURNAMENT_SIZE,
-        num_threads: int | None = None,
         elitism_size: int = DEFAULT_ELITISM_SIZE,
         adaptive_mutation: bool = DEFAULT_ADAPTIVE_MUTATION,
         enable_moo: bool = DEFAULT_ENABLE_MOO,
         enable_llm_crossover: bool = DEFAULT_ENABLE_LLM_CROSSOVER,
-        seed: int | None = DEFAULT_SEED,
         output_style_guidance: str | None = None,
         infer_output_style: bool = False,
-        verbose: int = 1,
         n_threads: int = DEFAULT_NUM_THREADS,
-        **model_kwargs: Any,
+        verbose: int = 1,
+        seed: int = DEFAULT_SEED,
     ) -> None:
-        """
-        Args:
-            model: The model to use for evaluation
-            population_size: Number of prompts in the population
-            num_generations: Number of generations to run
-            mutation_rate: Mutation rate for genetic operations
-            crossover_rate: Crossover rate for genetic operations
-            tournament_size: Tournament size for selection
-            n_threads: Number of threads for parallel evaluation
-            elitism_size: Number of elitism prompts
-            adaptive_mutation: Whether to use adaptive mutation
-            enable_moo: Whether to enable multi-objective optimization - When enable optimizes for both the supplied metric and the length of the prompt
-            enable_llm_crossover: Whether to enable LLM crossover
-            seed: Random seed for reproducibility
-            output_style_guidance: Output style guidance for prompts
-            infer_output_style: Whether to infer output style
-            verbose: Controls internal logging/progress bars (0=off, 1=on).
-            **model_kwargs: Additional model parameters
-        """
         # Initialize base class first
         if sys.version_info >= (3, 13):
             warnings.warn(
@@ -142,42 +143,27 @@ class EvolutionaryOptimizer(BaseOptimizer):
                 "You may see asyncio teardown warnings. Prefer Python 3.12.",
                 RuntimeWarning,
             )
-        if "project_name" in model_kwargs:
-            warnings.warn(
-                "The 'project_name' parameter in optimizer constructor is deprecated. "
-                "Set project_name in the ChatPrompt instead.",
-                DeprecationWarning,
-                stacklevel=2,
-            )
-            del model_kwargs["project_name"]
 
-        super().__init__(model=model, verbose=verbose, **model_kwargs)
+        super().__init__(
+            model=model, verbose=verbose, seed=seed, model_parameters=model_parameters
+        )
         self.population_size = population_size
         self.num_generations = num_generations
         self.mutation_rate = mutation_rate
         self.crossover_rate = crossover_rate
         self.tournament_size = tournament_size
-        if num_threads is not None:
-            warnings.warn(
-                "The 'num_threads' parameter is deprecated and will be removed in a future version. "
-                "Use 'n_threads' instead.",
-                DeprecationWarning,
-                stacklevel=2,
-            )
-            n_threads = num_threads
-        self.num_threads = n_threads
+        self.n_threads = n_threads
         self.elitism_size = elitism_size
         self.adaptive_mutation = adaptive_mutation
         self.enable_moo = enable_moo
         self.enable_llm_crossover = enable_llm_crossover
-        self.seed = seed if seed is not None else self.DEFAULT_SEED
+        self.seed = seed
         self.output_style_guidance = (
             output_style_guidance
             if output_style_guidance is not None
             else self.DEFAULT_OUTPUT_STYLE_GUIDANCE
         )
         self.infer_output_style = infer_output_style
-        self._current_optimization_id: str | None = None
         self._current_generation = 0
         self._best_fitness_history: list[float] = []
         self._generations_without_improvement = 0
@@ -249,8 +235,8 @@ class EvolutionaryOptimizer(BaseOptimizer):
                 func = getattr(cls, name)
                 setattr(self, name, func.__get__(self, self.__class__))
 
-        # LLM calls
-        bind(LlmSupport, ["_call_model"])
+        # LLM calls - now inherited from BaseOptimizer
+        # bind(LlmSupport, ["_call_model"])  # Removed - using BaseOptimizer._call_model
 
         # Mutations
         bind(
@@ -318,6 +304,7 @@ class EvolutionaryOptimizer(BaseOptimizer):
     ) -> Any:
         individual = creator.Individual(prompt_candidate.get_messages())
         setattr(individual, "tools", copy.deepcopy(prompt_candidate.tools))
+        setattr(individual, "function_map", prompt_candidate.function_map)
         return individual
 
     def _update_individual_with_prompt(
@@ -325,6 +312,7 @@ class EvolutionaryOptimizer(BaseOptimizer):
     ) -> Any:
         individual[:] = prompt_candidate.get_messages()
         setattr(individual, "tools", copy.deepcopy(prompt_candidate.tools))
+        setattr(individual, "function_map", prompt_candidate.function_map)
         return individual
 
     def _get_adaptive_mutation_rate(self) -> float:
@@ -415,13 +403,17 @@ class EvolutionaryOptimizer(BaseOptimizer):
         else:
             elites = tools.selBest(population, self.elitism_size)
 
-        seed_prompt = (
-            chat_prompt.ChatPrompt(
-                messages=max(elites, key=lambda x: x.fitness.values[0])
+        if elites:
+            best_elite = max(elites, key=lambda x: x.fitness.values[0])
+            seed_prompt = chat_prompt.ChatPrompt(
+                messages=best_elite,
+                tools=getattr(best_elite, "tools", best_prompt_so_far.tools),
+                function_map=getattr(
+                    best_elite, "function_map", best_prompt_so_far.function_map
+                ),
             )
-            if elites
-            else best_prompt_so_far
-        )
+        else:
+            seed_prompt = best_prompt_so_far
 
         prompt_variants = self._initialize_population(seed_prompt)
         new_pop = [self._create_individual_from_prompt(p) for p in prompt_variants]
@@ -522,6 +514,10 @@ class EvolutionaryOptimizer(BaseOptimizer):
         n_samples: int | None = None,
         auto_continue: bool = False,
         agent_class: type[OptimizableAgent] | None = None,
+        project_name: str = "Optimization",
+        max_trials: int = 10,
+        mcp_config: MCPExecutionConfig | None = None,
+        *args: Any,
         **kwargs: Any,
     ) -> OptimizationResult:
         """
@@ -533,21 +529,18 @@ class EvolutionaryOptimizer(BaseOptimizer):
             n_samples: Optional number of samples to use
             auto_continue: Whether to automatically continue optimization
             agent_class: Optional agent class to use
-            **kwargs: Additional keyword arguments including:
-                mcp_config (MCPExecutionConfig | None): MCP tool calling configuration (default: None)
+            project_name: Opik project name for logging traces (default: "Optimization")
+            mcp_config: MCP tool calling configuration (default: None)
         """
         # Use base class validation and setup methods
-        self.validate_optimization_inputs(prompt, dataset, metric)
-        self.configure_prompt_model(prompt)
-        self.agent_class = self.setup_agent_class(prompt, agent_class)
-
-        # Extract MCP config from kwargs (for optional MCP workflows)
-        mcp_config = kwargs.pop("mcp_config", None)
+        self._validate_optimization_inputs(prompt, dataset, metric)
+        self.agent_class = self._setup_agent_class(prompt, agent_class)
         evaluation_kwargs: dict[str, Any] = {}
         if mcp_config is not None:
             evaluation_kwargs["mcp_config"] = mcp_config
 
-        self.project_name = self.agent_class.project_name
+        # Set project name from parameter
+        self.project_name = project_name
 
         # Step 0. Start Opik optimization run
         opik_optimization_run: optimization.Optimization | None = None
@@ -557,14 +550,14 @@ class EvolutionaryOptimizer(BaseOptimizer):
                 objective_name=metric.__name__,
                 metadata={"optimizer": self.__class__.__name__},
             )
-            self._current_optimization_id = opik_optimization_run.id
+            self.current_optimization_id = opik_optimization_run.id
         except Exception as e:
             logger.warning(f"Opik server error: {e}. Continuing without Opik tracking.")
-            self._current_optimization_id = None
+            self.current_optimization_id = None
 
         reporting.display_header(
             algorithm=self.__class__.__name__,
-            optimization_id=self._current_optimization_id,
+            optimization_id=self.current_optimization_id,
             dataset_id=dataset.id,
             verbose=self.verbose,
         )
@@ -583,7 +576,8 @@ class EvolutionaryOptimizer(BaseOptimizer):
         )
 
         # Step 1. Step variables and define fitness function
-        self.reset_counters()  # Reset counters for run
+        self._reset_counters()  # Reset counters for run
+        trials_used = [0]  # Use list for closure mutability
         self._history: list[OptimizationRound] = []
         self._current_generation = 0
         self._best_fitness_history = []
@@ -595,7 +589,16 @@ class EvolutionaryOptimizer(BaseOptimizer):
 
             def _deap_evaluate_individual_fitness(
                 messages: list[dict[str, str]],
-            ) -> tuple[float, float]:
+            ) -> tuple[float, ...]:
+                # Check if we've hit the limit
+                if trials_used[0] >= max_trials:
+                    logger.debug(
+                        f"Skipping evaluation - max_trials ({max_trials}) reached"
+                    )
+                    return (-float("inf"), float("inf"))  # Worst possible fitness
+
+                trials_used[0] += 1
+
                 primary_fitness_score: float = self._evaluate_prompt(
                     prompt,
                     messages,  # type: ignore
@@ -603,7 +606,7 @@ class EvolutionaryOptimizer(BaseOptimizer):
                     metric=metric,
                     n_samples=n_samples,
                     experiment_config=(experiment_config or {}).copy(),
-                    optimization_id=self._current_optimization_id,
+                    optimization_id=self.current_optimization_id,
                     verbose=0,
                     **evaluation_kwargs,
                 )
@@ -614,7 +617,16 @@ class EvolutionaryOptimizer(BaseOptimizer):
             # Single-objective
             def _deap_evaluate_individual_fitness(
                 messages: list[dict[str, str]],
-            ) -> tuple[float, float]:
+            ) -> tuple[float, ...]:
+                # Check if we've hit the limit
+                if trials_used[0] >= max_trials:
+                    logger.debug(
+                        f"Skipping evaluation - max_trials ({max_trials}) reached"
+                    )
+                    return (-float("inf"),)  # Worst possible fitness
+
+                trials_used[0] += 1
+
                 fitness_score: float = self._evaluate_prompt(
                     prompt,
                     messages,  # type: ignore
@@ -622,11 +634,11 @@ class EvolutionaryOptimizer(BaseOptimizer):
                     metric=metric,
                     n_samples=n_samples,
                     experiment_config=(experiment_config or {}).copy(),
-                    optimization_id=self._current_optimization_id,
+                    optimization_id=self.current_optimization_id,
                     verbose=0,
                     **evaluation_kwargs,
                 )
-                return (fitness_score, 0.0)
+                return (fitness_score,)
 
         self.toolbox.register("evaluate", _deap_evaluate_individual_fitness)
 
@@ -644,6 +656,7 @@ class EvolutionaryOptimizer(BaseOptimizer):
                 else float(len(json.dumps(prompt.get_messages())))
             )
 
+            trials_used[0] = 0
             best_primary_score_overall = initial_primary_score
             best_prompt_overall = prompt
             report_baseline_performance.set_score(initial_primary_score)
@@ -715,14 +728,22 @@ class EvolutionaryOptimizer(BaseOptimizer):
                 )
                 best_primary_score_overall = current_best_for_primary.fitness.values[0]
                 best_prompt_overall = chat_prompt.ChatPrompt(
-                    messages=current_best_for_primary
+                    messages=current_best_for_primary,
+                    tools=getattr(current_best_for_primary, "tools", prompt.tools),
+                    function_map=getattr(
+                        current_best_for_primary, "function_map", prompt.function_map
+                    ),
                 )
             else:
                 # Single-objective
                 current_best_on_front = hof[0]
                 best_primary_score_overall = current_best_on_front.fitness.values[0]
                 best_prompt_overall = chat_prompt.ChatPrompt(
-                    messages=current_best_on_front
+                    messages=current_best_on_front,
+                    tools=getattr(current_best_on_front, "tools", prompt.tools),
+                    function_map=getattr(
+                        current_best_on_front, "function_map", prompt.function_map
+                    ),
                 )
 
             if self.enable_moo:
@@ -756,6 +777,13 @@ class EvolutionaryOptimizer(BaseOptimizer):
             verbose=self.verbose
         ) as report_evolutionary_algo:
             for generation_idx in range(1, self.num_generations + 1):
+                # Check if we've exhausted our evaluation budget
+                if trials_used[0] >= max_trials:
+                    logger.info(
+                        f"Stopping optimization: max_trials ({max_trials}) reached after {generation_idx - 1} generations"
+                    )
+                    break
+
                 report_evolutionary_algo.start_gen(generation_idx, self.num_generations)
 
                 curr_best_score = self._population_best_score(deap_population)
@@ -854,7 +882,11 @@ class EvolutionaryOptimizer(BaseOptimizer):
                     final_results_log += f"  Solution {i + 1}: Primary Score={sol.fitness.values[0]:.4f}, Length={sol.fitness.values[1]:.0f}, Prompt='{str(sol)[:100]}...'\n"
                 best_overall_solution = sorted_hof[0]
                 final_best_prompt = chat_prompt.ChatPrompt(
-                    messages=best_overall_solution
+                    messages=best_overall_solution,
+                    tools=getattr(best_overall_solution, "tools", prompt.tools),
+                    function_map=getattr(
+                        best_overall_solution, "function_map", prompt.function_map
+                    ),
                 )
                 final_primary_score = best_overall_solution.fitness.values[0]
                 final_length = best_overall_solution.fitness.values[1]
@@ -922,11 +954,12 @@ class EvolutionaryOptimizer(BaseOptimizer):
             )
 
         logger.info(f"Total LLM calls during optimization: {self.llm_call_counter}")
+        logger.info(f"Total prompt evaluations: {trials_used[0]}")
         if opik_optimization_run:
             try:
                 opik_optimization_run.update(status="completed")
                 logger.info(
-                    f"Opik Optimization run {self._current_optimization_id} status updated to completed."
+                    f"Opik Optimization run {self.current_optimization_id} status updated to completed."
                 )
             except Exception as e:
                 logger.warning(f"Failed to update Opik Optimization run status: {e}")
@@ -952,13 +985,14 @@ class EvolutionaryOptimizer(BaseOptimizer):
                 "seed": self.seed,
                 "prompt_type": "single_string_ga",
                 "initial_score_for_display": initial_score_for_display,
-                "temperature": self.model_kwargs.get("temperature"),
+                "temperature": self.model_parameters.get("temperature"),
                 "stopped_early": stopped_early_flag,
                 "rounds": self.get_history(),
                 "user_output_style_guidance": self.output_style_guidance,
                 "infer_output_style_requested": self.infer_output_style,
                 "final_effective_output_style_guidance": effective_output_style_guidance,
                 "infer_output_style": self.infer_output_style,
+                "trials_used": trials_used[0],
             }
         )
 
@@ -974,14 +1008,7 @@ class EvolutionaryOptimizer(BaseOptimizer):
         final_tools = getattr(final_best_prompt, "tools", None)
         if final_tools:
             final_details["final_tools"] = final_tools
-            tool_prompts = {
-                (tool.get("function", {}).get("name") or f"tool_{idx}"): tool.get(
-                    "function", {}
-                ).get("description")
-                for idx, tool in enumerate(final_tools)
-            }
-        else:
-            tool_prompts = None
+        tool_prompts = self._extract_tool_prompts(final_tools)
 
         return OptimizationResult(
             optimizer=self.__class__.__name__,
@@ -995,7 +1022,7 @@ class EvolutionaryOptimizer(BaseOptimizer):
             llm_calls=self.llm_call_counter,
             tool_calls=self.tool_call_counter,
             dataset_id=dataset.id,
-            optimization_id=self._current_optimization_id,
+            optimization_id=self.current_optimization_id,
             tool_prompts=tool_prompts,
         )
 
@@ -1090,7 +1117,7 @@ class EvolutionaryOptimizer(BaseOptimizer):
             self._mcp_context = previous_context
             self.enable_llm_crossover = previous_crossover
 
-        finalize_mcp_result(result, context, panel_style)
+        finalize_mcp_result(result, context, panel_style, optimizer=self)
         return result
 
     # Evaluation is provided by EvaluationOps
@@ -1102,7 +1129,7 @@ class EvolutionaryOptimizer(BaseOptimizer):
     def _get_reasoning_system_prompt_for_variation(self) -> str:
         return evo_prompts.variation_system_prompt(self.output_style_guidance)
 
-    def get_llm_crossover_system_prompt(self) -> str:
+    def _get_llm_crossover_system_prompt(self) -> str:
         return evo_prompts.llm_crossover_system_prompt(self.output_style_guidance)
 
     def _get_radical_innovation_system_prompt(self) -> str:

opik_optimizer/evolutionary_optimizer/mcp.py

@@ -168,16 +168,21 @@ def finalize_mcp_result(
     result: Any,
     context: EvolutionaryMCPContext,
     panel_style: str,
+    optimizer: Any = None,
 ) -> None:
     final_tools = (
         result.details.get("final_tools") if isinstance(result.details, dict) else None
     )
-    tool_prompts = {
-        (tool.get("function", {}).get("name") or tool.get("name")): tool.get(
-            "function", {}
-        ).get("description")
-        for tool in (final_tools or [])
-    }
+    # Use optimizer's centralized method if available, otherwise inline
+    if optimizer and hasattr(optimizer, "_extract_tool_prompts"):
+        tool_prompts = optimizer._extract_tool_prompts(final_tools) or {}
+    else:
+        tool_prompts = {
+            (tool.get("function", {}).get("name") or tool.get("name")): tool.get(
+                "function", {}
+            ).get("description")
+            for tool in (final_tools or [])
+        }
     if tool_prompts.get(context.tool_name):
         reporting.display_tool_description(
             tool_prompts[context.tool_name],

opik_optimizer/evolutionary_optimizer/mutation_ops.py

@@ -30,7 +30,11 @@ class MutationOps:
         self, individual: Any, initial_prompt: chat_prompt.ChatPrompt
     ) -> Any:
         """Enhanced mutation operation with multiple strategies."""
-        prompt = chat_prompt.ChatPrompt(messages=individual)
+        prompt = chat_prompt.ChatPrompt(
+            messages=individual,
+            tools=initial_prompt.tools,
+            function_map=initial_prompt.function_map,
+        )
 
         mcp_context = getattr(self, "_mcp_context", None)
         if mcp_context is not None:
@@ -130,7 +134,11 @@ class MutationOps:
                     f"Error parsing semantic mutation response as JSON. "
                     f"Response: {response!r}\nOriginal error: {parse_exc}"
                 ) from parse_exc
-            return chat_prompt.ChatPrompt(messages=messages)
+            return chat_prompt.ChatPrompt(
+                messages=messages,
+                tools=prompt.tools,
+                function_map=prompt.function_map,
+            )
         except Exception as e:
             reporting.display_error(
                 f"      Error in semantic mutation, this is usually a parsing error: {e}",
@@ -187,7 +195,11 @@ class MutationOps:
                 else:
                     mutated_messages.append({"role": role, "content": content})
 
-        return chat_prompt.ChatPrompt(messages=mutated_messages)
+        return chat_prompt.ChatPrompt(
+            messages=mutated_messages,
+            tools=prompt.tools,
+            function_map=prompt.function_map,
+        )
 
     def _word_level_mutation_prompt(
         self, prompt: chat_prompt.ChatPrompt
@@ -200,7 +212,11 @@ class MutationOps:
                     "content": self._word_level_mutation(message["content"]),
                 }
             )
-        return chat_prompt.ChatPrompt(messages=mutated_messages)
+        return chat_prompt.ChatPrompt(
+            messages=mutated_messages,
+            tools=prompt.tools,
+            function_map=prompt.function_map,
+        )
 
     def _word_level_mutation(self, msg_content: str) -> str:
         """Perform word-level mutation."""
@@ -298,7 +314,11 @@ class MutationOps:
                     f"Failed to parse LLM output in radical innovation mutation for prompt '{json.dumps(prompt.get_messages())[:50]}...'. Output: {new_prompt_str[:200]}. Error: {parse_exc}. Returning original."
                 )
                 return prompt
-            return chat_prompt.ChatPrompt(messages=new_messages)
+            return chat_prompt.ChatPrompt(
+                messages=new_messages,
+                tools=prompt.tools,
+                function_map=prompt.function_map,
+            )
         except Exception as e:
             logger.warning(
                 f"Radical innovation mutation failed for prompt '{json.dumps(prompt.get_messages())[:50]}...': {e}. Returning original."

opik_optimizer/evolutionary_optimizer/population_ops.py

@@ -88,13 +88,21 @@ class PopulationOps:
                             p.get("role") is not None for p in fresh_prompts
                         ):
                             population.append(
-                                chat_prompt.ChatPrompt(messages=fresh_prompts)
+                                chat_prompt.ChatPrompt(
+                                    messages=fresh_prompts,
+                                    tools=prompt.tools,
+                                    function_map=prompt.function_map,
+                                )
                             )
                             init_pop_report.success_fresh_prompts(1)
                         elif all(isinstance(p, list) for p in fresh_prompts):
                             population.extend(
                                 [
-                                    chat_prompt.ChatPrompt(messages=p)
+                                    chat_prompt.ChatPrompt(
+                                        messages=p,
+                                        tools=prompt.tools,
+                                        function_map=prompt.function_map,
+                                    )
                                     for p in fresh_prompts[:num_fresh_starts]
                                 ]
                             )
@@ -157,7 +165,11 @@ class PopulationOps:
                         )
                         population.extend(
                             [
-                                chat_prompt.ChatPrompt(messages=p)
+                                chat_prompt.ChatPrompt(
+                                    messages=p,
+                                    tools=prompt.tools,
+                                    function_map=prompt.function_map,
+                                )
                                 for p in generated_prompts_variations[
                                     :num_variations_on_initial
                                 ]
@@ -210,16 +222,20 @@ class PopulationOps:
         else:
             elites = tools.selBest(population, self.elitism_size)
 
-        seed_prompt = (
-            chat_prompt.ChatPrompt(
-                messages=max(elites, key=lambda x: x.fitness.values[0])
+        if elites:
+            best_elite = max(elites, key=lambda x: x.fitness.values[0])
+            seed_prompt = chat_prompt.ChatPrompt(
+                messages=best_elite,
+                tools=getattr(best_elite, "tools", best_prompt_so_far.tools),
+                function_map=getattr(
+                    best_elite, "function_map", best_prompt_so_far.function_map
+                ),
             )
-            if elites
-            else best_prompt_so_far
-        )
+        else:
+            seed_prompt = best_prompt_so_far
 
         prompt_variants = self._initialize_population(seed_prompt)
-        new_pop = [creator.Individual(p.get_messages()) for p in prompt_variants]
+        new_pop = [self._create_individual_from_prompt(p) for p in prompt_variants]  # type: ignore[attr-defined]
 
         for ind, fit in zip(new_pop, map(self.toolbox.evaluate, new_pop)):
             ind.fitness.values = fit