fix(llm): retry LiteLLM bad gateway errors (#12117)

openhands/llm/llm.py

@@ -21,6 +21,7 @@ from litellm import completion as litellm_completion
 from litellm import completion_cost as litellm_completion_cost
 from litellm.exceptions import (
     APIConnectionError,
+    BadGatewayError,
     RateLimitError,
     ServiceUnavailableError,
 )
@@ -45,6 +46,7 @@ LLM_RETRY_EXCEPTIONS: tuple[type[Exception], ...] = (
     APIConnectionError,
     RateLimitError,
     ServiceUnavailableError,
+    BadGatewayError,
     litellm.Timeout,
     litellm.InternalServerError,
     LLMNoResponseError,

tests/unit/llm/test_api_connection_error_retry.py

@@ -1,7 +1,7 @@
 from unittest.mock import patch
 
 import pytest
-from litellm.exceptions import APIConnectionError
+from litellm.exceptions import APIConnectionError, BadGatewayError
 
 from openhands.core.config import LLMConfig
 from openhands.llm.llm import LLM
@@ -86,3 +86,25 @@ def test_completion_max_retries_api_connection_error(
     # The exception doesn't contain retry information in the current implementation
     # Just verify that we got an APIConnectionError
     assert 'API connection error' in str(excinfo.value)
+
+
+@patch('openhands.llm.llm.litellm_completion')
+def test_completion_retries_bad_gateway_error(mock_litellm_completion, default_config):
+    """Test that BadGatewayError is properly retried."""
+    mock_litellm_completion.side_effect = [
+        BadGatewayError(
+            message='Bad gateway',
+            llm_provider='test_provider',
+            model='test_model',
+        ),
+        {'choices': [{'message': {'content': 'Retry successful'}}]},
+    ]
+
+    llm = LLM(config=default_config, service_id='test-service')
+    response = llm.completion(
+        messages=[{'role': 'user', 'content': 'Hello!'}],
+        stream=False,
+    )
+
+    assert response['choices'][0]['message']['content'] == 'Retry successful'
+    assert mock_litellm_completion.call_count == 2