feat: enable custom LLM support for Crew.test()

- Add llm parameter to Crew.test() that accepts string or LLM instance
- Maintain backward compatibility with openai_model_name parameter
- Update CrewEvaluator to handle any LLM implementation
- Add comprehensive test coverage

Fixes #2079

Co-Authored-By: Joe Moura <joao@crewai.com>

src/crewai/crew.py

@@ -1076,18 +1076,32 @@ class Crew(BaseModel):
         self,
         n_iterations: int,
         openai_model_name: Optional[str] = None,
+        llm: Optional[Union[str, LLM]] = None,
         inputs: Optional[Dict[str, Any]] = None,
     ) -> None:
-        """Test and evaluate the Crew with the given inputs for n iterations concurrently using concurrent.futures."""
+        """Test and evaluate the Crew with the given inputs for n iterations.
+
+        Args:
+            n_iterations: Number of test iterations to run
+            openai_model_name: (Deprecated) Name of OpenAI model to use for evaluation
+            llm: LLM instance or model name to use for evaluation
+            inputs: Optional dictionary of inputs to pass to the crew
+        """
+        if not llm and not openai_model_name:
+            raise ValueError("Either llm or openai_model_name must be provided")
+        
         test_crew = self.copy()
+        model = llm if llm else openai_model_name
+        if isinstance(model, str):
+            model = LLM(model=model)
 
         self._test_execution_span = test_crew._telemetry.test_execution_span(
             test_crew,
             n_iterations,
             inputs,
-            openai_model_name,  # type: ignore[arg-type]
+            str(model),  # type: ignore[arg-type]
         )  # type: ignore[arg-type]
-        evaluator = CrewEvaluator(test_crew, openai_model_name)  # type: ignore[arg-type]
+        evaluator = CrewEvaluator(test_crew, model)
 
         for i in range(1, n_iterations + 1):
             evaluator.set_iteration(i)

src/crewai/utilities/evaluators/crew_evaluator_handler.py

@@ -1,6 +1,10 @@
 from collections import defaultdict
 
+from typing import Union
+
 from pydantic import BaseModel, Field
+
+from crewai.llm import LLM
 from rich.box import HEAVY_EDGE
 from rich.console import Console
 from rich.table import Table
@@ -23,7 +27,7 @@ class CrewEvaluator:
 
     Attributes:
         crew (Crew): The crew of agents to evaluate.
-        openai_model_name (str): The model to use for evaluating the performance of the agents (for now ONLY OpenAI accepted).
+        llm (LLM): The language model to use for evaluating the performance of the agents.
         tasks_scores (defaultdict): A dictionary to store the scores of the agents for each task.
         iteration (int): The current iteration of the evaluation.
     """
@@ -32,9 +36,9 @@ class CrewEvaluator:
     run_execution_times: defaultdict = defaultdict(list)
     iteration: int = 0
 
-    def __init__(self, crew, openai_model_name: str):
+    def __init__(self, crew, llm: Union[str, LLM]):
         self.crew = crew
-        self.openai_model_name = openai_model_name
+        self.llm = llm if isinstance(llm, LLM) else LLM(model=llm)
         self._telemetry = Telemetry()
         self._setup_for_evaluating()
 
@@ -51,7 +55,7 @@ class CrewEvaluator:
             ),
             backstory="Evaluator agent for crew evaluation with precise capabilities to evaluate the performance of the agents in the crew based on the tasks they have performed",
             verbose=False,
-            llm=self.openai_model_name,
+            llm=self.llm,
         )
 
     def _evaluation_task(
@@ -181,7 +185,7 @@ class CrewEvaluator:
                 self.crew,
                 evaluation_result.pydantic.quality,
                 current_task._execution_time,
-                self.openai_model_name,
+                str(self.llm),
             )
             self.tasks_scores[self.iteration].append(evaluation_result.pydantic.quality)
             self.run_execution_times[self.iteration].append(

tests/crew_test.py

@@ -13,6 +13,7 @@ import pytest
 from crewai.agent import Agent
 from crewai.agents.cache import CacheHandler
 from crewai.crew import Crew
+from crewai.llm import LLM
 from crewai.crews.crew_output import CrewOutput
 from crewai.memory.contextual.contextual_memory import ContextualMemory
 from crewai.process import Process
@@ -1123,7 +1124,7 @@ def test_kickoff_for_each_empty_input():
     assert results == []
 
 
-@pytest.mark.vcr(filter_headers=["authorization"])
+@pytest.mark.vcr(filter_headeruvs=["authorization"])
 def test_kickoff_for_each_invalid_input():
     """Tests if kickoff_for_each raises TypeError for invalid input types."""
 
@@ -2812,10 +2813,43 @@ def test_conditional_should_execute():
 @mock.patch("crewai.crew.CrewEvaluator")
 @mock.patch("crewai.crew.Crew.copy")
 @mock.patch("crewai.crew.Crew.kickoff")
-def test_crew_testing_function(kickoff_mock, copy_mock, crew_evaluator):
+def test_crew_testing_with_custom_llm(kickoff_mock, copy_mock, crew_evaluator):
     task = Task(
-        description="Come up with a list of 5 interesting ideas to explore for an article, then write one amazing paragraph highlight for each idea that showcases how good an article about this topic could be. Return the list of ideas with their paragraph and your notes.",
-        expected_output="5 bullet points with a paragraph for each idea.",
+        description="Test task",
+        expected_output="Test output",
+        agent=researcher,
+    )
+
+    crew = Crew(
+        agents=[researcher],
+        tasks=[task],
+    )
+
+    # Create a mock for the copied crew
+    copy_mock.return_value = crew
+
+    custom_llm = LLM(model="gpt-4o-mini")
+    n_iterations = 2
+    crew.test(n_iterations, llm=custom_llm)
+
+    # Ensure kickoff is called on the copied crew
+    kickoff_mock.assert_has_calls([mock.call(inputs=None), mock.call(inputs=None)])
+
+    # Verify CrewEvaluator was called with custom LLM
+    crew_evaluator.assert_has_calls([
+        mock.call(crew, custom_llm),
+        mock.call().set_iteration(1),
+        mock.call().set_iteration(2),
+        mock.call().print_crew_evaluation_result(),
+    ])
+
+@mock.patch("crewai.crew.CrewEvaluator")
+@mock.patch("crewai.crew.Crew.copy")
+@mock.patch("crewai.crew.Crew.kickoff")
+def test_crew_testing_backward_compatibility(kickoff_mock, copy_mock, crew_evaluator):
+    task = Task(
+        description="Test task",
+        expected_output="Test output",
         agent=researcher,
     )
 
@@ -2831,19 +2865,41 @@ def test_crew_testing_function(kickoff_mock, copy_mock, crew_evaluator):
     crew.test(n_iterations, openai_model_name="gpt-4o-mini", inputs={"topic": "AI"})
 
     # Ensure kickoff is called on the copied crew
-    kickoff_mock.assert_has_calls(
-        [mock.call(inputs={"topic": "AI"}), mock.call(inputs={"topic": "AI"})]
+    kickoff_mock.assert_has_calls([
+        mock.call(inputs={"topic": "AI"}),
+        mock.call(inputs={"topic": "AI"})
+    ])
+
+    # Verify CrewEvaluator was called with string model name
+    crew_evaluator.assert_has_calls([
+        mock.call(crew, mock.ANY),
+        mock.call().set_iteration(1),
+        mock.call().set_iteration(2),
+        mock.call().print_crew_evaluation_result(),
+    ])
+
+@mock.patch("crewai.crew.CrewEvaluator")
+@mock.patch("crewai.crew.Crew.copy")
+@mock.patch("crewai.crew.Crew.kickoff")
+def test_crew_testing_missing_llm(kickoff_mock, copy_mock, crew_evaluator):
+    task = Task(
+        description="Test task",
+        expected_output="Test output",
+        agent=researcher,
     )
 
-    crew_evaluator.assert_has_calls(
-        [
-            mock.call(crew, "gpt-4o-mini"),
-            mock.call().set_iteration(1),
-            mock.call().set_iteration(2),
-            mock.call().print_crew_evaluation_result(),
-        ]
+    crew = Crew(
+        agents=[researcher],
+        tasks=[task],
     )
 
+    # Create a mock for the copied crew
+    copy_mock.return_value = crew
+
+    n_iterations = 2
+    with pytest.raises(ValueError, match="Either llm or openai_model_name must be provided"):
+        crew.test(n_iterations)
+
 
 @pytest.mark.vcr(filter_headers=["authorization"])
 def test_hierarchical_verbose_manager_agent():
@@ -3125,4 +3181,4 @@ def test_multimodal_agent_live_image_analysis():
     # Verify we got a meaningful response
     assert isinstance(result.raw, str)
     assert len(result.raw) > 100  # Expecting a detailed analysis
-    assert "error" not in result.raw.lower()  # No error messages in response
+    assert "error" not in result.raw.lower()  # No error messages in response

tests/utilities/evaluators/test_crew_evaluator_handler.py

@@ -23,7 +23,7 @@ class TestCrewEvaluator:
         )
         crew = Crew(agents=[agent], tasks=[task])
 
-        return CrewEvaluator(crew, openai_model_name="gpt-4o-mini")
+        return CrewEvaluator(crew, llm="gpt-4o-mini")
 
     def test_setup_for_evaluating(self, crew_planner):
         crew_planner._setup_for_evaluating()