fix(proxy): gate include_fallback_errors behind expose_fallback_errors_to_caller setting

Without an operator gate, any authenticated caller could set include_fallback_errors=True,
trigger a fallback, and read raw upstream exception messages from the
x-litellm-fallback-errors header and the litellm-fallback-metadata SSE event.

Strip include_fallback_errors from request data in common_processing_pre_call_logic
when expose_fallback_errors_to_caller is not set, so the router never builds the
error list. Also gate _should_include_fallback_errors on the same setting as a
secondary check for the streaming SSE injection path.

litellm/proxy/common_request_processing.py

@@ -1037,6 +1037,8 @@ class ProxyBaseLLMRequestProcessing:
             version=version,
             proxy_config=proxy_config,
         )
+        if not general_settings.get("expose_fallback_errors_to_caller"):
+            self.data.pop("include_fallback_errors", None)
         if route_type in {"aresponses", "_aresponses_websocket"}:
             await _authorize_response_file_search_vector_stores(
                 data=self.data,

litellm/proxy/proxy_server.py

@@ -7097,6 +7097,8 @@ def _is_positive_int_like(value: Any) -> bool:
 
 
 def _should_include_fallback_errors(request_data: dict[str, object]) -> bool:
+    if not general_settings.get("expose_fallback_errors_to_caller"):
+        return False
     return request_data.get("include_fallback_errors") is True
 
 

tests/test_litellm/test_router_streaming_fallback_metadata.py

@@ -4,6 +4,7 @@ from unittest.mock import MagicMock
 import pytest
 
 import litellm
+from litellm.proxy.proxy_server import _should_include_fallback_errors
 from litellm.router import Router
 from litellm.router_utils.add_retry_fallback_headers import get_hidden_params_dict
 
@@ -162,3 +163,25 @@ async def test_set_response_headers_adds_model_group_to_streaming_wrapper():
         "x-existing": "keep",
         "x-litellm-model-group": "fallback-model",
     }
+
+
+def test_should_include_fallback_errors_gated_by_operator_setting():
+    request_data: dict = {"include_fallback_errors": True}
+
+    import litellm.proxy.proxy_server as ps
+
+    original = ps.general_settings.copy() if isinstance(ps.general_settings, dict) else {}
+    try:
+        ps.general_settings = {}
+        assert _should_include_fallback_errors(request_data) is False
+
+        ps.general_settings = {"expose_fallback_errors_to_caller": False}
+        assert _should_include_fallback_errors(request_data) is False
+
+        ps.general_settings = {"expose_fallback_errors_to_caller": True}
+        assert _should_include_fallback_errors(request_data) is True
+
+        ps.general_settings = {"expose_fallback_errors_to_caller": True}
+        assert _should_include_fallback_errors({}) is False
+    finally:
+        ps.general_settings = original