fix: preserve original status code in handle_exception_on_proxy (#23901)

Root cause: handle_exception_on_proxy always maps non-HTTPException/non-ProxyException errors to 500, even when the original exception carries a status_code (e.g. 429 for rate limits).

Made-with: Cursor

litellm/proxy/utils.py

@@ -5148,11 +5148,17 @@ def handle_exception_on_proxy(e: Exception) -> ProxyException:
         )
     elif isinstance(e, ProxyException):
         return e
+    original_status_code = getattr(e, "status_code", None)
+    if isinstance(original_status_code, int) and 400 <= original_status_code < 600:
+        error_code = original_status_code
+    else:
+        error_code = status.HTTP_500_INTERNAL_SERVER_ERROR
     return ProxyException(
-        message="Internal Server Error, " + str(e),
+        message=getattr(e, "message", str(e)),
         type=ProxyErrorTypes.internal_server_error,
         param=getattr(e, "param", "None"),
-        code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+        code=error_code,
+        headers=getattr(e, "headers", None),
     )
 
 

tests/test_litellm/proxy/test_proxy_utils.py

@@ -190,3 +190,26 @@ def test_get_projected_spend_over_limit_includes_current_spend(monkeypatch):
     projected_spend, projected_exceeded_date = result
     assert projected_spend == 290.0
     assert projected_exceeded_date == real_datetime.date(2026, 4, 21)
+
+
+def test_handle_exception_on_proxy_preserves_429_status_code():
+    """Exceptions with status_code=429 should not be mapped to 500."""
+    from litellm.proxy.utils import handle_exception_on_proxy
+
+    class FakeRateLimitError(Exception):
+        def __init__(self):
+            self.status_code = 429
+            self.message = "Rate limit exceeded"
+            self.param = None
+            super().__init__(self.message)
+
+    result = handle_exception_on_proxy(FakeRateLimitError())
+    assert result.code == "429"
+
+
+def test_handle_exception_on_proxy_defaults_to_500_without_status_code():
+    """Exceptions without status_code should map to 500."""
+    from litellm.proxy.utils import handle_exception_on_proxy
+
+    result = handle_exception_on_proxy(Exception("unknown error"))
+    assert result.code == "500"