Address PR feedback: improve comments, maintain alphabetical order, and parameterize tests

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

src/crewai/llm.py

@@ -88,6 +88,7 @@ LLM_CONTEXT_WINDOW_SIZES = {
     "o1-preview": 128000,
     "o1-mini": 128000,
     "o3-mini": 200000,  # Based on official o3-mini specifications
+    "o4-mini": 200000,  # Token limit based on OpenAI's official specifications
     # gemini
     "gemini-2.0-flash": 1048576,
     "gemini-2.0-flash-thinking-exp-01-21": 32768,

src/crewai/tools/tool_usage.py

@@ -37,10 +37,11 @@ OPENAI_BIGGER_MODELS = [
     "gpt-4",
     "gpt-4o",
     "o1-preview",
-    "o1-mini",
     "o1",
+    "o1-mini",
     "o3",
     "o3-mini",
+    "o4-mini",  # Maintain alphabetical order for clarity
 ]
 
 

tests/agent_test.py

@@ -413,7 +413,9 @@ def test_agent_execution_with_specific_tools():
 
 
 @pytest.mark.vcr(filter_headers=["authorization"])
-def test_agent_powered_by_new_o_model_family_that_allows_skipping_tool():
+@pytest.mark.parametrize("model", ["o3-mini", "o4-mini"])
+def test_agent_powered_by_o_model_that_allows_skipping_tool(model):
+    """Test that o-series models can skip using tools when appropriate."""
     @tool
     def multiplier(first_number: int, second_number: int) -> float:
         """Useful for when you need to multiply two numbers together."""
@@ -423,7 +425,7 @@ def test_agent_powered_by_new_o_model_family_that_allows_skipping_tool():
         role="test role",
         goal="test goal",
         backstory="test backstory",
-        llm=LLM(model="o3-mini"),
+        llm=LLM(model=model),
         max_iter=3,
         use_system_prompt=False,
         allow_delegation=False,
@@ -439,9 +441,11 @@ def test_agent_powered_by_new_o_model_family_that_allows_skipping_tool():
 
 
 @pytest.mark.vcr(filter_headers=["authorization"])
-def test_agent_powered_by_new_o_model_family_that_uses_tool():
+@pytest.mark.parametrize("model", ["o3-mini", "o4-mini"])
+def test_agent_powered_by_o_model_that_uses_tool(model):
+    """Test that o-series models can use tools when appropriate."""
     @tool
-    def comapny_customer_data() -> float:
+    def company_customer_data() -> float:
         """Useful for getting customer related data."""
         return "The company has 42 customers"
 
@@ -449,7 +453,7 @@ def test_agent_powered_by_new_o_model_family_that_uses_tool():
         role="test role",
         goal="test goal",
         backstory="test backstory",
-        llm="o3-mini",
+        llm=model,
         max_iter=3,
         use_system_prompt=False,
         allow_delegation=False,
@@ -460,7 +464,7 @@ def test_agent_powered_by_new_o_model_family_that_uses_tool():
         agent=agent,
         expected_output="The number of customers",
     )
-    output = agent.execute_task(task=task, tools=[comapny_customer_data])
+    output = agent.execute_task(task=task, tools=[company_customer_data])
     assert output == "42"
 
 

tests/llm_test.py

@@ -335,12 +335,15 @@ def test_o3_mini_reasoning_effort_medium():
     assert "Paris" in result
 
 
-def test_context_window_validation():
+@pytest.mark.parametrize("model", ["o3-mini", "o4-mini"])
-    """Test that context window validation works correctly."""
+def test_context_window_validation_for_o_models(model):
-    # Test valid window size
+    """Test that context window validation works correctly for o-series models."""
-    llm = LLM(model="o3-mini")
+    # Test valid window size for o-series models
+    llm = LLM(model=model)
     assert llm.get_context_window_size() == int(200000 * CONTEXT_WINDOW_USAGE_RATIO)
 
+def test_context_window_validation():
+    """Test that context window validation works correctly."""
     # Test invalid window size
     with pytest.raises(ValueError) as excinfo:
         with patch.dict(