Fix import sorting in test file using ruff

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

src/crewai/agents/crew_agent_executor.py

@@ -232,7 +232,14 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
                 self._format_answer(answer)
             except OutputParserException as e:
                 if FINAL_ANSWER_AND_PARSABLE_ACTION_ERROR_MESSAGE in e.error:
-                    answer = answer.split("Observation:")[0].strip()
+                    # If both Action and Final Answer are present, prioritize the Action
+                    # by removing the Final Answer part
+                    if "Final Answer:" in answer:
+                        parts = answer.split("Final Answer:")
+                        answer = parts[0].strip()
+                    # If that doesn't work, try splitting at Observation
+                    elif "Observation:" in answer:
+                        answer = answer.split("Observation:")[0].strip()
 
         return self._format_answer(answer)
 

src/crewai/cli/constants.py

@@ -216,10 +216,43 @@ MODELS = {
         "watsonx/ibm/granite-3-8b-instruct",
     ],
     "bedrock": [
+        "bedrock/us.amazon.nova-pro-v1:0",
+        "bedrock/us.amazon.nova-micro-v1:0",
+        "bedrock/us.amazon.nova-lite-v1:0",
+        "bedrock/us.anthropic.claude-3-5-sonnet-20240620-v1:0",
+        "bedrock/us.anthropic.claude-3-5-haiku-20241022-v1:0",
+        "bedrock/us.anthropic.claude-3-5-sonnet-20241022-v2:0",
+        "bedrock/us.anthropic.claude-3-7-sonnet-20250219-v1:0",
+        "bedrock/us.anthropic.claude-3-sonnet-20240229-v1:0",
+        "bedrock/us.anthropic.claude-3-opus-20240229-v1:0",
+        "bedrock/us.anthropic.claude-3-haiku-20240307-v1:0",
+        "bedrock/us.meta.llama3-2-11b-instruct-v1:0",
+        "bedrock/us.meta.llama3-2-3b-instruct-v1:0",
+        "bedrock/us.meta.llama3-2-90b-instruct-v1:0",
+        "bedrock/us.meta.llama3-2-1b-instruct-v1:0",
+        "bedrock/us.meta.llama3-1-8b-instruct-v1:0",
+        "bedrock/us.meta.llama3-1-70b-instruct-v1:0",
+        "bedrock/us.meta.llama3-3-70b-instruct-v1:0",
+        "bedrock/us.meta.llama3-1-405b-instruct-v1:0",
+        "bedrock/eu.anthropic.claude-3-5-sonnet-20240620-v1:0",
+        "bedrock/eu.anthropic.claude-3-sonnet-20240229-v1:0",
+        "bedrock/eu.anthropic.claude-3-haiku-20240307-v1:0",
+        "bedrock/eu.meta.llama3-2-3b-instruct-v1:0",
+        "bedrock/eu.meta.llama3-2-1b-instruct-v1:0",
+        "bedrock/apac.anthropic.claude-3-5-sonnet-20240620-v1:0",
+        "bedrock/apac.anthropic.claude-3-5-sonnet-20241022-v2:0",
+        "bedrock/apac.anthropic.claude-3-sonnet-20240229-v1:0",
+        "bedrock/apac.anthropic.claude-3-haiku-20240307-v1:0",
+        "bedrock/amazon.nova-pro-v1:0",
+        "bedrock/amazon.nova-micro-v1:0",
+        "bedrock/amazon.nova-lite-v1:0",
         "bedrock/anthropic.claude-3-5-sonnet-20240620-v1:0",
+        "bedrock/anthropic.claude-3-5-haiku-20241022-v1:0",
+        "bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0",
+        "bedrock/anthropic.claude-3-7-sonnet-20250219-v1:0",
         "bedrock/anthropic.claude-3-sonnet-20240229-v1:0",
-        "bedrock/anthropic.claude-3-haiku-20240307-v1:0",
         "bedrock/anthropic.claude-3-opus-20240229-v1:0",
+        "bedrock/anthropic.claude-3-haiku-20240307-v1:0",
         "bedrock/anthropic.claude-v2:1",
         "bedrock/anthropic.claude-v2",
         "bedrock/anthropic.claude-instant-v1",
@@ -234,8 +267,6 @@ MODELS = {
         "bedrock/ai21.j2-mid-v1",
         "bedrock/ai21.j2-ultra-v1",
         "bedrock/ai21.jamba-instruct-v1:0",
-        "bedrock/meta.llama2-13b-chat-v1",
-        "bedrock/meta.llama2-70b-chat-v1",
         "bedrock/mistral.mistral-7b-instruct-v0:2",
         "bedrock/mistral.mixtral-8x7b-instruct-v0:1",
     ],

src/crewai/llm.py

@@ -64,6 +64,7 @@ LLM_CONTEXT_WINDOW_SIZES = {
     "gpt-4-turbo": 128000,
     "o1-preview": 128000,
     "o1-mini": 128000,
+    "o3-mini": 200000,  # Based on official o3-mini specifications
     # gemini
     "gemini-2.0-flash": 1048576,
     "gemini-1.5-pro": 2097152,
@@ -485,10 +486,23 @@ class LLM:
         """
         Returns the context window size, using 75% of the maximum to avoid
         cutting off messages mid-thread.
+
+        Raises:
+            ValueError: If a model's context window size is outside valid bounds (1024-2097152)
         """
         if self.context_window_size != 0:
             return self.context_window_size
 
+        MIN_CONTEXT = 1024
+        MAX_CONTEXT = 2097152  # Current max from gemini-1.5-pro
+
+        # Validate all context window sizes
+        for key, value in LLM_CONTEXT_WINDOW_SIZES.items():
+            if value < MIN_CONTEXT or value > MAX_CONTEXT:
+                raise ValueError(
+                    f"Context window for {key} must be between {MIN_CONTEXT} and {MAX_CONTEXT}"
+                )
+
         self.context_window_size = int(
             DEFAULT_CONTEXT_WINDOW_SIZE * CONTEXT_WINDOW_USAGE_RATIO
         )

tests/agents/test_action_final_answer_error.py

@@ -0,0 +1,84 @@
+from unittest.mock import MagicMock
+
+import pytest
+
+from crewai.agents.crew_agent_executor import CrewAgentExecutor
+from crewai.agents.parser import (
+    FINAL_ANSWER_AND_PARSABLE_ACTION_ERROR_MESSAGE,
+    AgentAction,
+    AgentFinish,
+    OutputParserException,
+)
+
+
+def test_process_llm_response_with_action_and_final_answer():
+    """Test that _process_llm_response correctly handles outputs with both Action and Final Answer."""
+    # Create a mock LLM
+    mock_llm = MagicMock()
+    mock_llm.supports_stop_words.return_value = False
+    
+    # Create a mock agent
+    mock_agent = MagicMock()
+    
+    # Create a CrewAgentExecutor instance
+    executor = CrewAgentExecutor(
+        llm=mock_llm,
+        task=MagicMock(),
+        crew=MagicMock(),
+        agent=mock_agent,
+        prompt={},
+        max_iter=5,
+        tools=[],
+        tools_names="",
+        stop_words=[],
+        tools_description="",
+        tools_handler=MagicMock(),
+    )
+    
+    # Test case 1: Output with both Action and Final Answer, with Final Answer after Action
+    output_with_both = """
+    Thought: I need to search for information and then provide an answer.
+    Action: search
+    Action Input: what is the temperature in SF?
+    Final Answer: The temperature is 100 degrees
+    """
+    
+    # Mock the _format_answer method to first raise an exception and then return a valid result
+    format_answer_mock = MagicMock()
+    format_answer_mock.side_effect = [
+        OutputParserException(FINAL_ANSWER_AND_PARSABLE_ACTION_ERROR_MESSAGE),
+        AgentAction(thought="", tool="search", tool_input="what is the temperature in SF?", text=""),
+    ]
+    executor._format_answer = format_answer_mock
+    
+    # Process the response
+    result = executor._process_llm_response(output_with_both)
+    
+    # Verify that the result is an AgentAction
+    assert isinstance(result, AgentAction)
+    assert result.tool == "search"
+    assert result.tool_input == "what is the temperature in SF?"
+    
+    # Test case 2: Output with both Action and Final Answer, with Observation in between
+    output_with_observation = """
+    Thought: I need to search for information.
+    Action: search
+    Action Input: what is the temperature in SF?
+    Observation: The temperature in SF is 100 degrees.
+    Final Answer: The temperature is 100 degrees
+    """
+    
+    # Reset the mock
+    format_answer_mock.reset_mock()
+    format_answer_mock.side_effect = [
+        OutputParserException(FINAL_ANSWER_AND_PARSABLE_ACTION_ERROR_MESSAGE),
+        AgentAction(thought="", tool="search", tool_input="what is the temperature in SF?", text=""),
+    ]
+    
+    # Process the response
+    result = executor._process_llm_response(output_with_observation)
+    
+    # Verify that the result is an AgentAction
+    assert isinstance(result, AgentAction)
+    assert result.tool == "search"
+    assert result.tool_input == "what is the temperature in SF?"

tests/llm_test.py

@@ -6,7 +6,7 @@ import pytest
 from pydantic import BaseModel
 
 from crewai.agents.agent_builder.utilities.base_token_process import TokenProcess
-from crewai.llm import LLM
+from crewai.llm import CONTEXT_WINDOW_USAGE_RATIO, LLM
 from crewai.utilities.events import crewai_event_bus
 from crewai.utilities.events.tool_usage_events import ToolExecutionErrorEvent
 from crewai.utilities.token_counter_callback import TokenCalcHandler
@@ -285,6 +285,23 @@ def test_o3_mini_reasoning_effort_medium():
     assert isinstance(result, str)
     assert "Paris" in result
 
+def test_context_window_validation():
+    """Test that context window validation works correctly."""
+    # Test valid window size
+    llm = LLM(model="o3-mini")
+    assert llm.get_context_window_size() == int(200000 * CONTEXT_WINDOW_USAGE_RATIO)
+
+    # Test invalid window size
+    with pytest.raises(ValueError) as excinfo:
+        with patch.dict(
+            "crewai.llm.LLM_CONTEXT_WINDOW_SIZES",
+            {"test-model": 500},  # Below minimum
+            clear=True,
+        ):
+            llm = LLM(model="test-model")
+            llm.get_context_window_size()
+    assert "must be between 1024 and 2097152" in str(excinfo.value)
+
 
 @pytest.mark.vcr(filter_headers=["authorization"])
 @pytest.fixture