fix: map Azure context_length_exceeded to ContextWindowExceededError

Azure Responses API returns a different error message format
('Your input exceeds the context window') than Chat Completions
('This model'\''s maximum context length is'). The existing check
only matched the latter, so Responses API context window errors
fell through to the generic 'invalid_request_error' handler and
were raised as BadRequestError.

Add checks for:
- azure_error_code == 'context_length_exceeded' (structured body)
- 'input exceeds the context window' (message string)

litellm/litellm_core_utils/exception_mapping_utils.py

@@ -2115,7 +2115,11 @@ def exception_type(  # type: ignore  # noqa: PLR0915
                         litellm_debug_info=extra_information,
                         response=getattr(original_exception, "response", None),
                     )
-                elif "This model's maximum context length is" in error_str:
+                elif (
+                    "This model's maximum context length is" in error_str
+                    or azure_error_code == "context_length_exceeded"
+                    or "input exceeds the context window" in error_str
+                ):
                     exception_mapping_worked = True
                     raise ContextWindowExceededError(
                         message=f"AzureException ContextWindowExceededError - {message}",

tests/test_litellm/llms/azure/test_azure_exception_mapping.py

@@ -10,7 +10,7 @@ sys.path.insert(
 )  # Adds the parent directory to the system path
 
 import litellm
-from litellm.exceptions import ContentPolicyViolationError
+from litellm.exceptions import ContentPolicyViolationError, ContextWindowExceededError
 from litellm.litellm_core_utils.exception_mapping_utils import exception_type
 
 
@@ -419,6 +419,37 @@ class TestAzureExceptionMapping:
         assert "optional_pre_call_checks" in error.message
         assert "docs.litellm.ai" in error.message
 
+    def test_azure_context_length_exceeded_via_error_code(self):
+        """Azure Responses API returns context_length_exceeded in the error code
+        with a different message format than the chat completions API. This
+        should still be mapped to ContextWindowExceededError."""
+
+        mock_exception = Exception(
+            "Your input exceeds the context window of this model. "
+            "Please adjust your input and try again."
+        )
+        mock_exception.body = {
+            "error": {
+                "message": (
+                    "Your input exceeds the context window of this model. "
+                    "Please adjust your input and try again."
+                ),
+                "type": "invalid_request_error",
+                "param": "input",
+                "code": "context_length_exceeded",
+            }
+        }
+        mock_response = MagicMock()
+        mock_response.status_code = 400
+        mock_exception.response = mock_response
+
+        with pytest.raises(ContextWindowExceededError):
+            exception_type(
+                model="azure/gpt-4o",
+                original_exception=mock_exception,
+                custom_llm_provider="azure",
+            )
+
     def test_openai_invalid_encrypted_content_error(self):
         """Test that OpenAI invalid_encrypted_content errors also get helpful guidance."""
         from litellm.exceptions import BadRequestError