From 1e02e695f2784fa594b9e30cb7e3805d625eb2d9 Mon Sep 17 00:00:00 2001
From: Jaeyeon Kim <anencore94@gmail.com>
Date: Sat, 28 Feb 2026 02:10:27 +0900
Subject: [PATCH] feat: add include_litellm_headers setting to strip
 x-litellm-* response headers

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---
 litellm/__init__.py                           |  1 +
 litellm/proxy/common_request_processing.py    |  4 ++
 .../proxy/test_common_request_processing.py   | 67 +++++++++++++++++++
 3 files changed, 72 insertions(+)

diff --git a/litellm/__init__.py b/litellm/__init__.py
index 50fa0e76755..8756c515680 100644
--- a/litellm/__init__.py
+++ b/litellm/__init__.py
@@ -351,6 +351,7 @@ anthropic_beta_headers_url: str = os.getenv(
     "https://raw.githubusercontent.com/BerriAI/litellm/main/litellm/anthropic_beta_headers_config.json",
 )
 suppress_debug_info = False
+include_litellm_headers: bool = True
 dynamodb_table_name: Optional[str] = None
 s3_callback_params: Optional[Dict] = None
 datadog_llm_observability_params: Optional[Union[DatadogLLMObsInitParams, Dict]] = None
diff --git a/litellm/proxy/common_request_processing.py b/litellm/proxy/common_request_processing.py
index 1269f58213a..19b18d1cba0 100644
--- a/litellm/proxy/common_request_processing.py
+++ b/litellm/proxy/common_request_processing.py
@@ -395,6 +395,10 @@ class ProxyBaseLLMRequestProcessing:
         litellm_logging_obj: Optional[LiteLLMLoggingObj] = None,
         **kwargs,
     ) -> dict:
+        # When include_litellm_headers is False, skip all x-litellm-* headers
+        if litellm.include_litellm_headers is False:
+            return {}
+
         exclude_values = {"", None, "None"}
         hidden_params = hidden_params or {}
 
diff --git a/tests/test_litellm/proxy/test_common_request_processing.py b/tests/test_litellm/proxy/test_common_request_processing.py
index 9b905d24fd1..84b614978cb 100644
--- a/tests/test_litellm/proxy/test_common_request_processing.py
+++ b/tests/test_litellm/proxy/test_common_request_processing.py
@@ -691,6 +691,73 @@ class TestProxyBaseLLMRequestProcessing:
             float(headers_7["x-litellm-key-spend"]) == 0.001
         )  # Should use original spend on error
 
+    def test_get_custom_headers_included_by_default(self):
+        """
+        Test that x-litellm-* headers are included by default
+        (backwards compatibility).
+        """
+        mock_user_api_key_dict = MagicMock(spec=UserAPIKeyAuth)
+        mock_user_api_key_dict.tpm_limit = None
+        mock_user_api_key_dict.rpm_limit = None
+        mock_user_api_key_dict.max_budget = None
+        mock_user_api_key_dict.spend = 0.0
+
+        # Ensure default is True
+        original_value = litellm.include_litellm_headers
+        litellm.include_litellm_headers = True
+        try:
+            headers = ProxyBaseLLMRequestProcessing.get_custom_headers(
+                user_api_key_dict=mock_user_api_key_dict,
+                call_id="test-call-id",
+                model_id="test-model-id",
+                version="1.0.0",
+            )
+
+            assert "x-litellm-call-id" in headers
+            assert headers["x-litellm-call-id"] == "test-call-id"
+            assert "x-litellm-model-id" in headers
+            assert headers["x-litellm-model-id"] == "test-model-id"
+            assert "x-litellm-version" in headers
+            assert headers["x-litellm-version"] == "1.0.0"
+        finally:
+            litellm.include_litellm_headers = original_value
+
+    def test_get_custom_headers_excluded_when_disabled(self):
+        """
+        Test that x-litellm-* headers are excluded when
+        include_litellm_headers is set to False.
+        """
+        mock_user_api_key_dict = MagicMock(spec=UserAPIKeyAuth)
+        mock_user_api_key_dict.tpm_limit = None
+        mock_user_api_key_dict.rpm_limit = None
+        mock_user_api_key_dict.max_budget = None
+        mock_user_api_key_dict.spend = 0.0
+
+        original_value = litellm.include_litellm_headers
+        litellm.include_litellm_headers = False
+        try:
+            headers = ProxyBaseLLMRequestProcessing.get_custom_headers(
+                user_api_key_dict=mock_user_api_key_dict,
+                call_id="test-call-id",
+                model_id="test-model-id",
+                version="1.0.0",
+                response_cost=0.001,
+                hidden_params={
+                    "_response_ms": 100.0,
+                    "litellm_overhead_time_ms": 10.0,
+                },
+            )
+
+            # All headers should be stripped
+            assert headers == {}
+            assert "x-litellm-call-id" not in headers
+            assert "x-litellm-model-id" not in headers
+            assert "x-litellm-version" not in headers
+            assert "x-litellm-response-cost" not in headers
+            assert "x-litellm-key-spend" not in headers
+        finally:
+            litellm.include_litellm_headers = original_value
+
     @pytest.mark.asyncio
     async def test_queue_time_seconds_is_set_in_metadata(self, monkeypatch):
         """