use copy to split testing and training on crews (#1491)

* use copy to split testing and training on crews

* make tests handle new copy functionality on train and test

* fix last test

* fix test

src/crewai/crew.py

@@ -435,15 +435,16 @@ def train(
         self, n_iterations: int, filename: str, inputs: Optional[Dict[str, Any]] = {}
     ) -> None:
         """Trains the crew for a given number of iterations."""
-        self._setup_for_training(filename)
+        train_crew = self.copy()
+        train_crew._setup_for_training(filename)
 
         for n_iteration in range(n_iterations):
-            self._train_iteration = n_iteration
-            self.kickoff(inputs=inputs)
+            train_crew._train_iteration = n_iteration
+            train_crew.kickoff(inputs=inputs)
 
         training_data = CrewTrainingHandler(TRAINING_DATA_FILE).load()
 
-        for agent in self.agents:
+        for agent in train_crew.agents:
             result = TaskEvaluator(agent).evaluate_training_data(
                 training_data=training_data, agent_id=str(agent.id)
             )
@@ -987,17 +988,19 @@ def test(
         inputs: Optional[Dict[str, Any]] = None,
     ) -> None:
         """Test and evaluate the Crew with the given inputs for n iterations concurrently using concurrent.futures."""
-        self._test_execution_span = self._telemetry.test_execution_span(
-            self,
+        test_crew = self.copy()
+
+        self._test_execution_span = test_crew._telemetry.test_execution_span(
+            test_crew,
             n_iterations,
             inputs,
             openai_model_name,  # type: ignore[arg-type]
         )  # type: ignore[arg-type]
-        evaluator = CrewEvaluator(self, openai_model_name)  # type: ignore[arg-type]
+        evaluator = CrewEvaluator(test_crew, openai_model_name)  # type: ignore[arg-type]
 
         for i in range(1, n_iterations + 1):
             evaluator.set_iteration(i)
-            self.kickoff(inputs=inputs)
+            test_crew.kickoff(inputs=inputs)
 
         evaluator.print_crew_evaluation_result()
 

tests/crew_test.py

@@ -9,6 +9,7 @@
 import instructor
 import pydantic_core
 import pytest
+
 from crewai.agent import Agent
 from crewai.agents.cache import CacheHandler
 from crewai.crew import Crew
@@ -497,6 +498,7 @@ def multiplier(first_number: int, second_number: int) -> float:
 @pytest.mark.vcr(filter_headers=["authorization"])
 def test_api_calls_throttling(capsys):
     from unittest.mock import patch
+
     from crewai_tools import tool
 
     @tool
@@ -779,11 +781,14 @@ def test_async_task_execution_call_count():
     list_important_history.output = mock_task_output
     write_article.output = mock_task_output
 
-    with patch.object(
-        Task, "execute_sync", return_value=mock_task_output
-    ) as mock_execute_sync, patch.object(
-        Task, "execute_async", return_value=mock_future
-    ) as mock_execute_async:
+    with (
+        patch.object(
+            Task, "execute_sync", return_value=mock_task_output
+        ) as mock_execute_sync,
+        patch.object(
+            Task, "execute_async", return_value=mock_future
+        ) as mock_execute_async,
+    ):
         crew.kickoff()
 
         assert mock_execute_async.call_count == 2
@@ -1105,6 +1110,7 @@ def crew_callback(_):
 @pytest.mark.vcr(filter_headers=["authorization"])
 def test_crew_function_calling_llm():
     from unittest.mock import patch
+
     from crewai_tools import tool
 
     llm = "gpt-4o"
@@ -1448,52 +1454,6 @@ def test_crew_does_not_interpolate_without_inputs():
             interpolate_task_inputs.assert_not_called()
 
 
-# def test_crew_partial_inputs():
-#     agent = Agent(
-#         role="{topic} Researcher",
-#         goal="Express hot takes on {topic}.",
-#         backstory="You have a lot of experience with {topic}.",
-#     )
-
-#     task = Task(
-#         description="Give me an analysis around {topic}.",
-#         expected_output="{points} bullet points about {topic}.",
-#     )
-
-#     crew = Crew(agents=[agent], tasks=[task], inputs={"topic": "AI"})
-#     inputs = {"topic": "AI"}
-#     crew._interpolate_inputs(inputs=inputs)  # Manual call for now
-
-#     assert crew.tasks[0].description == "Give me an analysis around AI."
-#     assert crew.tasks[0].expected_output == "{points} bullet points about AI."
-#     assert crew.agents[0].role == "AI Researcher"
-#     assert crew.agents[0].goal == "Express hot takes on AI."
-#     assert crew.agents[0].backstory == "You have a lot of experience with AI."
-
-
-# def test_crew_invalid_inputs():
-#     agent = Agent(
-#         role="{topic} Researcher",
-#         goal="Express hot takes on {topic}.",
-#         backstory="You have a lot of experience with {topic}.",
-#     )
-
-#     task = Task(
-#         description="Give me an analysis around {topic}.",
-#         expected_output="{points} bullet points about {topic}.",
-#     )
-
-#     crew = Crew(agents=[agent], tasks=[task], inputs={"subject": "AI"})
-#     inputs = {"subject": "AI"}
-#     crew._interpolate_inputs(inputs=inputs)  # Manual call for now
-
-#     assert crew.tasks[0].description == "Give me an analysis around {topic}."
-#     assert crew.tasks[0].expected_output == "{points} bullet points about {topic}."
-#     assert crew.agents[0].role == "{topic} Researcher"
-#     assert crew.agents[0].goal == "Express hot takes on {topic}."
-#     assert crew.agents[0].backstory == "You have a lot of experience with {topic}."
-
-
 def test_task_callback_on_crew():
     from unittest.mock import MagicMock, patch
 
@@ -1770,7 +1730,10 @@ def testing_tool(first_number: int, second_number: int) -> int:
 @patch("crewai.crew.Crew.kickoff")
 @patch("crewai.crew.CrewTrainingHandler")
 @patch("crewai.crew.TaskEvaluator")
-def test_crew_train_success(task_evaluator, crew_training_handler, kickoff):
+@patch("crewai.crew.Crew.copy")
+def test_crew_train_success(
+    copy_mock, task_evaluator, crew_training_handler, kickoff_mock
+):
     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.",
@@ -1781,9 +1744,19 @@ def test_crew_train_success(task_evaluator, crew_training_handler, kickoff):
         agents=[researcher, writer],
         tasks=[task],
     )
+
+    # Create a mock for the copied crew
+    copy_mock.return_value = crew
+
     crew.train(
         n_iterations=2, inputs={"topic": "AI"}, filename="trained_agents_data.pkl"
     )
+
+    # Ensure kickoff is called on the copied crew
+    kickoff_mock.assert_has_calls(
+        [mock.call(inputs={"topic": "AI"}), mock.call(inputs={"topic": "AI"})]
+    )
+
     task_evaluator.assert_has_calls(
         [
             mock.call(researcher),
@@ -1822,10 +1795,6 @@ def test_crew_train_success(task_evaluator, crew_training_handler, kickoff):
         ]
     )
 
-    kickoff.assert_has_calls(
-        [mock.call(inputs={"topic": "AI"}), mock.call(inputs={"topic": "AI"})]
-    )
-
 
 def test_crew_train_error():
     task = Task(
@@ -1840,7 +1809,7 @@ def test_crew_train_error():
     )
 
     with pytest.raises(TypeError) as e:
-        crew.train()
+        crew.train()  # type: ignore purposefully throwing err
         assert "train() missing 1 required positional argument: 'n_iterations'" in str(
             e
         )
@@ -2536,8 +2505,9 @@ 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(mock_kickoff, crew_evaluator):
+def test_crew_testing_function(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.",
@@ -2548,11 +2518,15 @@ def test_crew_testing_function(mock_kickoff, crew_evaluator):
         agents=[researcher],
         tasks=[task],
     )
+
+    # Create a mock for the copied crew
+    copy_mock.return_value = crew
+
     n_iterations = 2
     crew.test(n_iterations, openai_model_name="gpt-4o-mini", inputs={"topic": "AI"})
 
-    assert len(mock_kickoff.mock_calls) == n_iterations
-    mock_kickoff.assert_has_calls(
+    # Ensure kickoff is called on the copied crew
+    kickoff_mock.assert_has_calls(
         [mock.call(inputs={"topic": "AI"}), mock.call(inputs={"topic": "AI"})]
     )