Add type hints, validation, and documentation to _create_manager_agent method

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

.gitignore

@@ -22,4 +22,6 @@ crew_tasks_output.json
 .ruff_cache
 .venv
 agentops.log
-test_flow.html
+test_flow.html# Test cassettes with sensitive data
+tests/cassettes/*_with_knowledge_sources.yaml
+tests/cassettes/*_sensitive_*.yaml

src/crewai/crew.py

@@ -735,7 +735,20 @@ class Crew(BaseModel):
         self._create_manager_agent()
         return self._execute_tasks(self.tasks)
 
-    def _create_manager_agent(self):
+    def _create_manager_agent(self) -> Agent:
+        """Create a manager agent for hierarchical process.
+        
+        Creates or configures a manager agent that will be responsible for delegating tasks
+        to other agents in a hierarchical process. If knowledge sources are provided,
+        they will be passed to the manager agent to enhance its context awareness.
+        
+        Returns:
+            Agent: The configured manager agent
+        
+        Raises:
+            Exception: If the manager agent has tools, which is not allowed
+            ValueError: If knowledge sources are provided but not valid BaseKnowledgeSource instances
+        """
         i18n = I18N(prompt_file=self.prompt_file)
         if self.manager_agent is not None:
             self.manager_agent.allow_delegation = True
@@ -748,6 +761,13 @@ class Crew(BaseModel):
                 raise Exception("Manager agent should not have tools")
         else:
             self.manager_llm = create_llm(self.manager_llm)
+            
+            # Validate knowledge sources if provided
+            if self.knowledge_sources and not all(
+                isinstance(ks, BaseKnowledgeSource) for ks in self.knowledge_sources
+            ):
+                raise ValueError("All knowledge sources must be instances of BaseKnowledgeSource")
+                
             manager = Agent(
                 role=i18n.retrieve("hierarchical_manager_agent", "role"),
                 goal=i18n.retrieve("hierarchical_manager_agent", "goal"),
@@ -755,10 +775,12 @@ class Crew(BaseModel):
                 tools=AgentTools(agents=self.agents).tools(),
                 allow_delegation=True,
                 llm=self.manager_llm,
+                knowledge_sources=self.knowledge_sources,
                 verbose=self.verbose,
             )
             self.manager_agent = manager
         manager.crew = self
+        return manager
 
     def _execute_tasks(
         self,

tests/agent_test.py

@@ -13,6 +13,7 @@ from crewai.agents.parser import AgentAction, CrewAgentParser, OutputParserExcep
 from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
 from crewai.knowledge.source.string_knowledge_source import StringKnowledgeSource
 from crewai.llm import LLM
+from crewai.process import Process
 from crewai.tools import tool
 from crewai.tools.tool_calling import InstructorToolCalling
 from crewai.tools.tool_usage import ToolUsage
@@ -1797,3 +1798,52 @@ def test_litellm_anthropic_error_handling():
 
     # Verify the LLM call was only made once (no retries)
     mock_llm_call.assert_called_once()
+
+
+@pytest.mark.vcr(filter_headers=["authorization"])
+def test_custom_llm_with_knowledge_sources():
+    """Test that knowledge sources work with custom LLMs in hierarchical crews."""
+    # Create a knowledge source with some content
+    content = "Brandon's favorite color is red and he likes Mexican food."
+    string_source = StringKnowledgeSource(content=content)
+
+    # Create a custom LLM
+    custom_llm = LLM(model="gpt-3.5-turbo")
+
+    with patch(
+        "crewai.knowledge.storage.knowledge_storage.KnowledgeStorage"
+    ) as MockKnowledge:
+        mock_knowledge_instance = MockKnowledge.return_value
+        mock_knowledge_instance.sources = [string_source]
+        mock_knowledge_instance.query.return_value = [{"content": content}]
+
+        # Create an agent with the custom LLM
+        agent = Agent(
+            role="Information Agent",
+            goal="Provide information based on knowledge sources",
+            backstory="You have access to specific knowledge sources.",
+            llm=custom_llm,
+        )
+
+        # Create a task that requires the agent to use the knowledge
+        task = Task(
+            description="What is Brandon's favorite color?",
+            expected_output="Brandon's favorite color.",
+            agent=agent,
+        )
+
+        # Create a crew with hierarchical process and custom LLM as manager
+        crew = Crew(
+            agents=[agent],
+            tasks=[task],
+            process=Process.hierarchical,
+            manager_llm=custom_llm,
+            knowledge_sources=[string_source],
+        )
+        
+        with patch.object(crew, "_execute_tasks") as mock_execute_tasks:
+            mock_execute_tasks.return_value.raw = "Brandon's favorite color is red."
+            result = crew.kickoff()
+
+            # Assert that the agent provides the correct information
+            assert "red" in result.raw.lower()