fix: respect custom trained_agents_data_file during inference

Agents always loaded from the hardcoded 'trained_agents_data.pkl' during
inference, ignoring any custom filename supplied at training time via
'crewai train -f <custom>.pkl'.

Changes:
- Add 'trained_agents_data_file' field to Crew (defaults to
  'trained_agents_data.pkl') so users can specify which file to load
  trained agent suggestions from during inference.
- Update Agent._use_trained_data() to accept an optional filename
  parameter instead of always using the hardcoded constant.
- Update apply_training_data() in agent/utils.py to propagate the
  crew's trained_agents_data_file to the agent.
- Add tests for custom filename propagation at agent and crew levels.

Closes #4905

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

lib/crewai/src/crewai/agent/core.py

@@ -1010,9 +1010,19 @@ class Agent(BaseAgent):
 
         return task_prompt
 
-    def _use_trained_data(self, task_prompt: str) -> str:
-        """Use trained data for the agent task prompt to improve output."""
-        if data := CrewTrainingHandler(TRAINED_AGENTS_DATA_FILE).load():
+    def _use_trained_data(
+        self, task_prompt: str, trained_agents_data_file: str | None = None
+    ) -> str:
+        """Use trained data for the agent task prompt to improve output.
+
+        Args:
+            task_prompt: The task prompt to augment.
+            trained_agents_data_file: Optional path to the trained agents data
+                file. Falls back to the default ``TRAINED_AGENTS_DATA_FILE``
+                when not provided.
+        """
+        filename = trained_agents_data_file or TRAINED_AGENTS_DATA_FILE
+        if data := CrewTrainingHandler(filename).load():
             if trained_data_output := data.get(self.role):
                 task_prompt += (
                     "\n\nYou MUST follow these instructions: \n - "

lib/crewai/src/crewai/agent/utils.py

@@ -222,7 +222,13 @@ def apply_training_data(agent: Agent, task_prompt: str) -> str:
     """
     if agent.crew and agent.crew._train:
         return agent._training_handler(task_prompt=task_prompt)
-    return agent._use_trained_data(task_prompt=task_prompt)
+    trained_agents_data_file = (
+        agent.crew.trained_agents_data_file if agent.crew else None
+    )
+    return agent._use_trained_data(
+        task_prompt=task_prompt,
+        trained_agents_data_file=trained_agents_data_file,
+    )
 
 
 def process_tool_results(agent: Agent, result: Any) -> Any:

lib/crewai/src/crewai/crew.py

@@ -96,7 +96,11 @@ from crewai.tools.agent_tools.read_file_tool import ReadFileTool
 from crewai.tools.base_tool import BaseTool
 from crewai.types.streaming import CrewStreamingOutput
 from crewai.types.usage_metrics import UsageMetrics
-from crewai.utilities.constants import NOT_SPECIFIED, TRAINING_DATA_FILE
+from crewai.utilities.constants import (
+    NOT_SPECIFIED,
+    TRAINED_AGENTS_DATA_FILE,
+    TRAINING_DATA_FILE,
+)
 from crewai.utilities.crew.models import CrewContext
 from crewai.utilities.evaluators.crew_evaluator_handler import CrewEvaluator
 from crewai.utilities.evaluators.task_evaluator import TaskEvaluator
@@ -303,6 +307,14 @@ class Crew(FlowTrackable, BaseModel):
         default=None,
         description="Whether to enable tracing for the crew. True=always enable, False=always disable, None=check environment/user settings.",
     )
+    trained_agents_data_file: str = Field(
+        default=TRAINED_AGENTS_DATA_FILE,
+        description=(
+            "Path to the file containing trained agent suggestions. "
+            "Defaults to 'trained_agents_data.pkl'. Set this to match the "
+            "custom filename used during training (e.g., via `crewai train -f`)."
+        ),
+    )
 
     @field_validator("id", mode="before")
     @classmethod

lib/crewai/tests/agents/test_agent.py

@@ -1064,6 +1064,59 @@ def test_agent_use_trained_data(crew_training_handler):
     )
 
 
+@patch("crewai.agent.core.CrewTrainingHandler")
+def test_agent_use_trained_data_with_custom_filename(crew_training_handler):
+    """Test that _use_trained_data respects a custom filename when provided."""
+    task_prompt = "What is 1 + 1?"
+    agent = Agent(
+        role="researcher",
+        goal="test goal",
+        backstory="test backstory",
+        verbose=True,
+    )
+    crew_training_handler.return_value.load.return_value = {
+        agent.role: {
+            "suggestions": [
+                "The result of the math operation must be right.",
+                "Result must be better than 1.",
+            ]
+        }
+    }
+
+    custom_filename = "my_custom_trained.pkl"
+    result = agent._use_trained_data(
+        task_prompt=task_prompt, trained_agents_data_file=custom_filename
+    )
+
+    assert (
+        result == "What is 1 + 1?\n\nYou MUST follow these instructions: \n"
+        " - The result of the math operation must be right.\n - Result must be better than 1."
+    )
+    crew_training_handler.assert_has_calls(
+        [mock.call(custom_filename), mock.call().load()]
+    )
+
+
+@patch("crewai.agent.core.CrewTrainingHandler")
+def test_agent_use_trained_data_defaults_without_custom_filename(crew_training_handler):
+    """Test that _use_trained_data falls back to the default file when no custom filename is given."""
+    task_prompt = "What is 1 + 1?"
+    agent = Agent(
+        role="researcher",
+        goal="test goal",
+        backstory="test backstory",
+        verbose=True,
+    )
+    crew_training_handler.return_value.load.return_value = {}
+
+    result = agent._use_trained_data(task_prompt=task_prompt)
+
+    assert result == task_prompt
+    crew_training_handler.assert_has_calls(
+        [mock.call("trained_agents_data.pkl"), mock.call().load()]
+    )
+
+
 def test_agent_max_retry_limit():
     agent = Agent(
         role="test role",

lib/crewai/tests/test_crew.py

@@ -2974,6 +2974,75 @@ def test__setup_for_training(researcher, writer):
         assert agent.allow_delegation is False
 
 
+def test_crew_trained_agents_data_file_defaults(researcher, writer):
+    """Test that Crew.trained_agents_data_file defaults to 'trained_agents_data.pkl'."""
+    task = Task(
+        description="Test task",
+        expected_output="Test output",
+        agent=researcher,
+    )
+    crew = Crew(agents=[researcher, writer], tasks=[task])
+    assert crew.trained_agents_data_file == "trained_agents_data.pkl"
+
+
+def test_crew_trained_agents_data_file_custom(researcher, writer):
+    """Test that Crew.trained_agents_data_file can be set to a custom value."""
+    task = Task(
+        description="Test task",
+        expected_output="Test output",
+        agent=researcher,
+    )
+    crew = Crew(
+        agents=[researcher, writer],
+        tasks=[task],
+        trained_agents_data_file="my_custom_trained.pkl",
+    )
+    assert crew.trained_agents_data_file == "my_custom_trained.pkl"
+
+
+@patch("crewai.agent.core.CrewTrainingHandler")
+def test_apply_training_data_uses_crew_custom_filename(mock_handler, researcher):
+    """Test that apply_training_data propagates the crew's trained_agents_data_file."""
+    from crewai.agent.utils import apply_training_data
+
+    task = Task(
+        description="Test task",
+        expected_output="Test output",
+        agent=researcher,
+    )
+    crew = Crew(
+        agents=[researcher],
+        tasks=[task],
+        trained_agents_data_file="my_custom_trained.pkl",
+    )
+    researcher.crew = crew
+
+    mock_handler.return_value.load.return_value = {
+        researcher.role: {
+            "suggestions": ["Be concise."]
+        }
+    }
+
+    result = apply_training_data(researcher, "Do the task")
+
+    mock_handler.assert_called_with("my_custom_trained.pkl")
+    assert "Be concise." in result
+
+
+@patch("crewai.agent.core.CrewTrainingHandler")
+def test_apply_training_data_uses_default_when_no_crew(mock_handler, researcher):
+    """Test that apply_training_data falls back to the default file when agent has no crew."""
+    from crewai.agent.utils import apply_training_data
+
+    researcher.crew = None
+    mock_handler.return_value.load.return_value = {}
+
+    result = apply_training_data(researcher, "Do the task")
+
+    mock_handler.assert_called_with("trained_agents_data.pkl")
+    assert result == "Do the task"
+
+
 @pytest.mark.vcr()
 def test_replay_feature(researcher, writer):
     list_ideas = Task(