feat: add context window size for o3-mini model (#2192)

* feat: add context window size for o3-mini model

Fixes #2191

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

* feat: add context window validation and tests

- Add validation for context window size bounds (1024-2097152)
- Add test for context window validation
- Fix test import error

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

* style: fix import sorting in llm_test.py

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

---------

Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: Joe Moura <joao@crewai.com>
Co-authored-by: Brandon Hancock (bhancock_ai) <109994880+bhancockio@users.noreply.github.com>

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/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