[Bug Fix] Exception mapping for context window exceeded - should catch vertex AI claude exceptions

litellm/litellm_core_utils/exception_mapping_utils.py

@@ -43,16 +43,16 @@ def is_error_str_rate_limit(error_str: str) -> bool:
         """
         if not isinstance(error_str, str):
             return False
-        
+
         if "429" in error_str or "rate limit" in error_str.lower():
             return True
-        
+
         #######################################
         # Mistral API returns this error string
         #########################################
         if "service tier capacity exceeded" in error_str.lower():
             return True
-        
+
         return False
 
     @staticmethod
@@ -68,6 +68,7 @@ def is_error_str_context_window_exceeded(error_str: str) -> bool:
             "model's maximum context limit",
             "is longer than the model's context length",
             "input tokens exceed the configured limit",
+            "input length and `max_tokens` exceed context limit",
         ]
         for substring in known_exception_substrings:
             if substring in _error_str_lowercase:

tests/test_litellm/litellm_core_utils/test_exception_mapping_utils.py

@@ -6,27 +6,39 @@
 # Tuple format: (error_message, expected_result)
 context_window_test_cases = [
     # Positive cases (should return True)
-    ("An error occurred: The input exceeds the model's maximum context limit of 8192 tokens.", True),
-    ("Some text before, this model's maximum context length is 4096 tokens. Some text after.", True),
-    ("Validation Error: string too long. expected a string with maximum length 1000.", True),
+    (
+        "An error occurred: The input exceeds the model's maximum context limit of 8192 tokens.",
+        True,
+    ),
+    (
+        "Some text before, this model's maximum context length is 4096 tokens. Some text after.",
+        True,
+    ),
+    (
+        "Validation Error: string too long. expected a string with maximum length 1000.",
+        True,
+    ),
     ("Your prompt is longer than the model's context length of 2048.", True),
     ("AWS Bedrock Error: The request payload size has exceed context limit.", True),
-    ("Input tokens exceed the configured limit of 272000 tokens. Your messages resulted in 509178 tokens. Please reduce the length of the messages.", True),
-
+    (
+        "Input tokens exceed the configured limit of 272000 tokens. Your messages resulted in 509178 tokens. Please reduce the length of the messages.",
+        True,
+    ),
+    ("input length and `max_tokens` exceed context limit", True),
     # Test case insensitivity
     ("ERROR: THIS MODEL'S MAXIMUM CONTEXT LENGTH IS 1024.", True),
-
     # Negative cases (should return False)
     ("A generic API error occurred.", False),
     ("Invalid API Key provided.", False),
     ("Rate limit reached for requests.", False),
     ("The context is large, but acceptable.", False),
-    ("", False), # Empty string
+    ("", False),  # Empty string
 ]
 
+
 @pytest.mark.parametrize("error_str, expected", context_window_test_cases)
 def test_is_error_str_context_window_exceeded(error_str, expected):
     """
     Tests the is_error_str_context_window_exceeded function with various error strings.
     """
-    assert ExceptionCheckers.is_error_str_context_window_exceeded(error_str) == expected
+    assert ExceptionCheckers.is_error_str_context_window_exceeded(error_str) == expected