fix(drop_params): warn when a deployment or env drop_params value is not a flag

A deployment drop_params string that is not a flag value (a typo like ture) stayed silently off. The router now logs one warning per deployment. LITELLM_DROP_PARAMS and litellm_settings.drop_params share the same helper, so a non-flag value there warns as well instead of flipping silently from on to off

litellm/__init__.py

@@ -47,7 +47,7 @@ from typing import (
 )
 from litellm.types.integrations.datadog import DatadogInitParams
 from litellm.types.integrations.newrelic import NewRelicInitParams
-from litellm.litellm_core_utils.core_helpers import normalize_drop_params
+from litellm.litellm_core_utils.core_helpers import drop_params_flag
 from litellm._logging import (
     set_verbose,
     _turn_on_debug,
@@ -239,7 +239,7 @@ token: Optional[str] = (
 )
 telemetry = True
 max_tokens: int = DEFAULT_MAX_TOKENS  # OpenAI Defaults
-drop_params = bool(normalize_drop_params(os.getenv("LITELLM_DROP_PARAMS")))
+drop_params = drop_params_flag(os.getenv("LITELLM_DROP_PARAMS"), "LITELLM_DROP_PARAMS", verbose_logger)
 modify_params = bool(os.getenv("LITELLM_MODIFY_PARAMS", False))
 use_chat_completions_url_for_anthropic_messages: bool = bool(
     os.getenv("LITELLM_USE_CHAT_COMPLETIONS_URL_FOR_ANTHROPIC_MESSAGES", False)

litellm/litellm_core_utils/core_helpers.py

@@ -1,6 +1,7 @@
 # What is this?
 ## Helper utilities
 import copy
+import logging
 from collections.abc import Iterable, Mapping
 from typing import TYPE_CHECKING, Any, Final, Literal
 
@@ -50,6 +51,13 @@ def normalize_drop_params(value: object) -> bool | None:
         return None
 
 
+def drop_params_flag(value: object, source: str, logger: logging.Logger) -> bool:
+    normalized: Final = normalize_drop_params(value)
+    if normalized is None and value is not None:
+        logger.warning("%s=%r is not a flag value, treating it as off", source, value)
+    return bool(normalized)
+
+
 def safe_divide(
     numerator: float,
     denominator: float,

litellm/proxy/proxy_server.py

@@ -278,8 +278,8 @@ from litellm.litellm_core_utils.asyncify import asyncify
 from litellm.litellm_core_utils.audio_utils.utils import resolve_speech_media_type
 from litellm.litellm_core_utils.core_helpers import (
     _get_parent_otel_span_from_kwargs,
+    drop_params_flag,
     get_litellm_metadata_from_kwargs,
-    normalize_drop_params,
 )
 from litellm.litellm_core_utils.credential_accessor import CredentialAccessor
 from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
@@ -5510,7 +5510,7 @@ class ProxyConfig:
                     parse_budget_reset_time(value)
                     setattr(litellm, key, value)
                 elif key == "drop_params":
-                    litellm.drop_params = _drop_params_from_litellm_settings(value)
+                    litellm.drop_params = drop_params_flag(value, "litellm_settings.drop_params", verbose_proxy_logger)
                 else:
                     verbose_proxy_logger.debug(
                         "%s setting litellm.%s=%s%s",
@@ -16915,13 +16915,6 @@ def _redact_config_param_value_for_logging(param_name: str | None, param_value:
     return param_value
 
 
-def _drop_params_from_litellm_settings(value: object) -> bool:
-    normalized: Final = normalize_drop_params(value)
-    if normalized is None and value is not None:
-        verbose_proxy_logger.warning("litellm_settings.drop_params=%r is not a flag value, treating it as off", value)
-    return bool(normalized)
-
-
 def _redact_general_setting_value(field_name: str, value: JsonValue, is_full_admin: bool) -> JsonValue:
     if is_full_admin:
         return value

litellm/router.py

@@ -9366,6 +9366,12 @@ class Router:
         #### VALIDATE MODEL ########
         # Check if this is a prompt management model before validating as LLM provider
         litellm_model: Final = deployment.litellm_params.model
+        if isinstance(deployment.litellm_params.drop_params, str):
+            verbose_router_logger.warning(
+                "model=%s drop_params=%r is not a flag value, treating it as unset",
+                deployment.model_name,
+                deployment.litellm_params.drop_params,
+            )
         is_prompt_management_model = False
 
         if "/" in litellm_model:

tests/test_litellm/litellm_core_utils/test_core_helpers.py

@@ -1,9 +1,12 @@
 """Tests for litellm_core_utils.core_helpers module."""
 
+import logging
+
 import pytest
 
 from litellm.litellm_core_utils.core_helpers import (
     _FINISH_REASON_MAP,
+    drop_params_flag,
     get_or_create_metadata_bucket,
     map_finish_reason,
     normalize_drop_params,
@@ -284,6 +287,20 @@ def test_normalize_drop_params(value, expected):
     assert normalize_drop_params(value) is expected
 
 
+@pytest.mark.parametrize("value, expected", [("true", True), ("off", False), (None, False)])
+def test_drop_params_flag_returns_a_bool_without_a_warning(value, expected, caplog):
+    with caplog.at_level(logging.WARNING, logger="drop-params-test"):
+        assert drop_params_flag(value, "LITELLM_DROP_PARAMS", logging.getLogger("drop-params-test")) is expected
+    assert caplog.text == ""
+
+
+@pytest.mark.parametrize("value", ["temperature", "ture", 2])
+def test_drop_params_flag_treats_non_flag_values_as_off_with_a_warning(value, caplog):
+    with caplog.at_level(logging.WARNING, logger="drop-params-test"):
+        assert drop_params_flag(value, "LITELLM_DROP_PARAMS", logging.getLogger("drop-params-test")) is False
+    assert f"LITELLM_DROP_PARAMS={value!r} is not a flag value, treating it as off" in caplog.text
+
+
 class TestIsExpectedClientError:
     def test_status_ranges(self):
         from litellm.litellm_core_utils.core_helpers import is_expected_client_error

tests/test_litellm/test_drop_params_env_var.py

@@ -5,13 +5,26 @@ import sys
 import pytest
 
 
-@pytest.mark.parametrize("configured, expected", [("false", "False"), ("true", "True")])
-def test_litellm_drop_params_env_var_is_parsed_as_a_flag(configured, expected):
-    result = subprocess.run(
+def _import_litellm_with(configured: str) -> subprocess.CompletedProcess[str]:
+    return subprocess.run(
         [sys.executable, "-c", "import litellm; print(litellm.drop_params)"],
         env={**os.environ, "LITELLM_DROP_PARAMS": configured},
         capture_output=True,
         text=True,
         check=True,
     )
+
+
+@pytest.mark.parametrize("configured, expected", [("false", "False"), ("true", "True")])
+def test_litellm_drop_params_env_var_is_parsed_as_a_flag(configured, expected):
+    result = _import_litellm_with(configured)
+
     assert result.stdout.strip() == expected
+    assert "is not a flag value" not in result.stderr
+
+
+def test_litellm_drop_params_env_var_non_flag_value_is_off_with_a_warning():
+    result = _import_litellm_with("temperature")
+
+    assert result.stdout.strip() == "False"
+    assert "LITELLM_DROP_PARAMS='temperature' is not a flag value, treating it as off" in result.stderr

tests/test_litellm/test_router.py

@@ -14424,3 +14424,36 @@ async def test_router_deployment_drop_params_string_true_is_honored(monkeypatch)
         temperature=0.1,
     )
     assert response.choices[0].message.content == "Hello, world!"
+
+
+@pytest.mark.parametrize("value", ["ture", "enabled"])
+def test_router_warns_when_a_deployment_drop_params_string_is_not_a_flag(value, caplog):
+    with caplog.at_level(logging.WARNING, logger="LiteLLM Router"):
+        router = Router(
+            model_list=[
+                {
+                    "model_name": "gpt-5-nano",
+                    "litellm_params": {"model": "openai/gpt-5-nano", "api_key": "sk-fake", "drop_params": value},
+                }
+            ]
+        )
+
+    deployment = router.get_deployment_by_model_group_name(model_group_name="gpt-5-nano")
+    assert deployment is not None
+    assert deployment.litellm_params.drop_params == value
+    assert f"model=gpt-5-nano drop_params={value!r} is not a flag value, treating it as unset" in caplog.text
+
+
+@pytest.mark.parametrize("value", [True, "true", "off", None])
+def test_router_stays_quiet_when_a_deployment_drop_params_is_a_flag(value, caplog):
+    with caplog.at_level(logging.WARNING, logger="LiteLLM Router"):
+        Router(
+            model_list=[
+                {
+                    "model_name": "gpt-5-nano",
+                    "litellm_params": {"model": "openai/gpt-5-nano", "api_key": "sk-fake", "drop_params": value},
+                }
+            ]
+        )
+
+    assert "is not a flag value" not in caplog.text