fix(http_handler): surface BaseLLMException raised while streaming a request body

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

litellm/llms/custom_httpx/llm_http_handler.py

@@ -44,7 +44,7 @@ from litellm.llms.base_llm.base_model_iterator import (
     MockResponseIterator,
 )
 from litellm.llms.base_llm.batches.transformation import BaseBatchesConfig
-from litellm.llms.base_llm.chat.transformation import BaseConfig
+from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
 from litellm.llms.base_llm.containers.transformation import BaseContainerConfig
 from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
 from litellm.llms.base_llm.evals.transformation import BaseEvalsAPIConfig
@@ -311,6 +311,24 @@ def _collect_ws_project_quota_callbacks() -> tuple[ProjectQuotaCallback, ...]:
     )
 
 
+def _find_base_llm_exception(error: BaseException, depth: int = 0) -> BaseLLMException | None:
+    if depth >= 10:
+        return None
+    if isinstance(error, BaseLLMException):
+        return error
+
+    cause: Final = error.__cause__
+    if cause is not None:
+        provider_exception: Final = _find_base_llm_exception(cause, depth + 1)
+        if provider_exception is not None:
+            return provider_exception
+
+    context: Final = error.__context__
+    if context is not None:
+        return _find_base_llm_exception(context, depth + 1)
+    return None
+
+
 class BaseLLMHTTPHandler:
     async def _make_common_async_call(
         self,
@@ -5980,6 +5998,10 @@ class BaseLLMHTTPHandler:
             BaseEvalsAPIConfig,
         ],
     ):
+        provider_exception: Final = _find_base_llm_exception(e)
+        if provider_exception is not None:
+            raise provider_exception
+
         received_status_code: Final = (
             e.response.status_code if isinstance(e, httpx.HTTPStatusError) else getattr(e, "status_code", None)
         )
@@ -6000,8 +6022,6 @@ class BaseLLMHTTPHandler:
             error_headers = {}
 
         if provider_config is None:
-            from litellm.llms.base_llm.chat.transformation import BaseLLMException
-
             raise BaseLLMException(
                 status_code=status_code,
                 message=error_text,

tests/test_litellm/llms/custom_httpx/test_llm_http_handler.py

@@ -2110,6 +2110,47 @@ def test_sync_retrieve_file_content_raises_on_http_error():
     assert exc_info.value.status_code == 404
 
 
+def test_handle_error_preserves_base_llm_exception_message_and_status():
+    from litellm.llms.vertex_ai.common_utils import VertexAIError
+    from litellm.llms.vertex_ai.files.transformation import VertexAIFilesConfig
+
+    message = "Invalid JSON on line 3 of batch input file: Expecting value"
+    error = VertexAIError(status_code=400, message=message)
+
+    with pytest.raises(VertexAIError) as exc_info:
+        BaseLLMHTTPHandler()._handle_error(e=error, provider_config=VertexAIFilesConfig())
+
+    assert exc_info.value.status_code == 400
+    assert exc_info.value.message == message
+
+
+def test_handle_error_preserves_base_llm_exception_from_wrapper():
+    from litellm.llms.vertex_ai.common_utils import VertexAIError
+    from litellm.llms.vertex_ai.files.transformation import VertexAIFilesConfig
+
+    message = "Invalid JSON on line 3 of batch input file: Expecting value"
+    error = VertexAIError(status_code=400, message=message)
+    wrapper = httpx.ConnectError("Failed to send bytes")
+    wrapper.__cause__ = error
+
+    with pytest.raises(VertexAIError) as exc_info:
+        BaseLLMHTTPHandler()._handle_error(e=wrapper, provider_config=VertexAIFilesConfig())
+
+    assert exc_info.value.status_code == 400
+    assert exc_info.value.message == message
+
+
+def test_handle_error_maps_plain_exception_to_provider_error():
+    from litellm.llms.vertex_ai.common_utils import VertexAIError
+    from litellm.llms.vertex_ai.files.transformation import VertexAIFilesConfig
+
+    with pytest.raises(VertexAIError) as exc_info:
+        BaseLLMHTTPHandler()._handle_error(e=ValueError("boom"), provider_config=VertexAIFilesConfig())
+
+    assert exc_info.value.status_code == 500
+    assert exc_info.value.message == "boom"
+
+
 _UPSTREAM_NOT_FOUND_BODY = {
     "error": {
         "message": "Response with id 'resp_abc' not found.",