fix(router): keep unresolved drop_params strings so DB rows and env refs survive

The drop_params validator collapsed every string it did not recognize to None. A pre-fix DB row holds the flag as ciphertext, so a partial PATCH rebuilt the deployment without it and dropped the key from the stored row, and /model/new turned an os.environ/ reference into nothing before the loader could resolve it. The validator now returns the raw value when it is not a boolean flag, the field admits strings the way timeout already does, and the flag set follows pydantic's lax bool parsing instead of a hand-rolled true/false pair

litellm/litellm_core_utils/core_helpers.py

@@ -5,6 +5,7 @@ from collections.abc import Iterable, Mapping
 from typing import TYPE_CHECKING, Any, Final, Literal
 
 import httpx
+from pydantic import TypeAdapter, ValidationError
 
 from litellm._logging import verbose_logger
 from litellm.types.llms.openai import AllMessageValues, OpenAIChatCompletionFinishReason
@@ -37,16 +38,16 @@ def safe_divide_seconds(seconds: float, denominator: float, default: float | Non
     return float(seconds / denominator)
 
 
+_DROP_PARAMS_BOOL: Final = TypeAdapter(bool)
+
+
 def normalize_drop_params(value: object) -> bool | None:
     if isinstance(value, bool):
         return value
-    if isinstance(value, str):
-        lowered: Final = value.strip().lower()
-        if lowered == "true":
-            return True
-        if lowered == "false":
-            return False
-    return None
+    try:
+        return _DROP_PARAMS_BOOL.validate_python(value.strip() if isinstance(value, str) else value)
+    except ValidationError:
+        return None
 
 
 def safe_divide(

litellm/types/router.py

@@ -315,7 +315,7 @@ class GenericLiteLLMParams(CredentialLiteLLMParams, CustomPricingLiteLLMParams):
     timeout: float | str | httpx.Timeout | None = None  # if str, pass in as os.environ/
     stream_timeout: float | str | None = None  # timeout when making stream=True calls, if str, pass in as os.environ/
     max_retries: int | None = None
-    drop_params: bool | None = None
+    drop_params: bool | str | None = None
     organization: str | None = None  # for openai orgs
     configurable_clientside_auth_params: CONFIGURABLE_CLIENTSIDE_AUTH_PARAMS = None
     litellm_credential_name: str | None = None
@@ -408,8 +408,9 @@ class GenericLiteLLMParams(CredentialLiteLLMParams, CustomPricingLiteLLMParams):
 
     @field_validator("drop_params", mode="before")
     @classmethod
-    def coerce_drop_params(cls, value: object) -> bool | None:
-        return normalize_drop_params(value)
+    def coerce_drop_params(cls, value: object) -> object:
+        normalized: Final = normalize_drop_params(value)
+        return value if normalized is None else normalized
 
     def __contains__(self, key) -> bool:
         # Define custom behavior for the 'in' operator

tests/test_litellm/litellm_core_utils/test_core_helpers.py

@@ -268,11 +268,16 @@ class TestRedactNestedMatchAndRegexKeys:
         (" TRUE ", True),
         ("false", False),
         ("False", False),
+        ("yes", True),
+        ("off", False),
+        ("1", True),
+        (1, True),
+        (0, False),
         (None, None),
-        ("yes", None),
         ("", None),
-        (1, None),
-        (0, None),
+        ("os.environ/DROP_PARAMS", None),
+        ("v2:gcm:not-a-flag", None),
+        (2, None),
     ],
 )
 def test_normalize_drop_params(value, expected):

tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py

@@ -30,7 +30,7 @@ from litellm.proxy.management_endpoints.model_management_endpoints import (
 )
 from litellm.proxy.utils import PrismaClient
 from litellm.router import Router
-from litellm.types.router import Deployment, LiteLLM_Params, updateDeployment
+from litellm.types.router import Deployment, LiteLLM_Params, ModelInfo, updateDeployment, updateLiteLLMParams
 
 
 async def _passthrough_row(update_data):
@@ -3070,6 +3070,31 @@ class TestUpdateDBModelBlocked:
         assert "blocked" not in result
 
 
+class TestUpdateDBModelKeepsLegacyDropParams:
+    def test_partial_patch_keeps_encrypted_string_drop_params(self, monkeypatch):
+        from litellm.proxy.common_utils.encrypt_decrypt_utils import decrypt_value_helper
+        from litellm.proxy.management_endpoints.model_management_endpoints import update_db_model
+
+        monkeypatch.setenv("LITELLM_SALT_KEY", "sk-1234")
+        legacy_row = Deployment(
+            model_name="gpt-5-nano",
+            litellm_params=LiteLLM_Params(
+                model="openai/gpt-5-nano",
+                api_key=encrypt_value_helper(value="sk-old"),
+                drop_params=encrypt_value_helper(value="true"),
+            ),
+            model_info=ModelInfo(id="legacy-row"),
+        )
+
+        result = update_db_model(
+            db_model=legacy_row,
+            updated_patch=updateDeployment(litellm_params=updateLiteLLMParams(api_key="sk-new")),
+        )
+
+        stored = json.loads(result["litellm_params"])
+        assert decrypt_value_helper(value=stored["drop_params"], key="drop_params") == "true"
+
+
 def _build_db_model_with_pricing():
     """Wildcard deployment with custom pricing in litellm_params; Deployment.__init__
     mirrors SPECIAL_MODEL_INFO_PARAMS into model_info, so both blobs hold the rate."""

tests/test_litellm/proxy/proxy_server/test_proxy_config.py

@@ -19,6 +19,7 @@ import pytest
 
 import litellm
 from litellm.proxy._types import CommonProxyErrors
+from litellm.proxy.common_utils.encrypt_decrypt_utils import encrypt_value_helper
 from litellm.proxy.proxy_server import (
     ProxyConfig,
     _is_remote_module_url,
@@ -2401,6 +2402,35 @@ def test_ProxyConfig__add_deployment_resolves_env_refs_on_arbitrary_field(monkey
     assert deployment.litellm_params.some_future_field == "resolved-custom-value"
 
 
+@pytest.mark.parametrize(
+    "stored_drop_params",
+    ["true", "os.environ/DROP_PARAMS_FLAG"],
+)
+def test_ProxyConfig__add_deployment_turns_stored_drop_params_string_into_bool(monkeypatch, stored_drop_params):
+    monkeypatch.setenv("LITELLM_SALT_KEY", "sk-1234")
+    monkeypatch.setenv("DROP_PARAMS_FLAG", "true")
+    fake_router = MagicMock()
+    fake_router.upsert_deployment = MagicMock(return_value=True)
+    monkeypatch.setattr("litellm.proxy.proxy_server.llm_router", fake_router)
+    pc = ProxyConfig()
+    db_model = SimpleNamespace(
+        model_id="model-1",
+        model_name="gpt-5-nano",
+        model_info={"id": "model-1"},
+        litellm_params={
+            "model": encrypt_value_helper(value="openai/gpt-5-nano"),
+            "drop_params": encrypt_value_helper(value=stored_drop_params),
+        },
+        blocked=False,
+    )
+
+    added = pc._add_deployment(db_models=[db_model])
+    deployment = fake_router.upsert_deployment.call_args.kwargs["deployment"]
+
+    assert added == 1
+    assert deployment.litellm_params.drop_params is True
+
+
 # ---------------------------------------------------------------------------
 # ProxyConfig.decrypt_model_list_from_db
 # ---------------------------------------------------------------------------

tests/test_litellm/types/test_router.py

@@ -1,8 +1,10 @@
 import pytest
+from pydantic import ValidationError
 
 from litellm.types.router import (
     SPECIAL_MODEL_INFO_PARAMS,
     Deployment,
+    GenericLiteLLMParams,
     LiteLLM_Params,
     ModelInfo,
 )
@@ -89,3 +91,24 @@ def test_pricing_strings_are_coerced_to_float():
 def test_invalid_pricing_is_rejected():
     with pytest.raises(ValueError, match='validation error for ModelInfo'):
         ModelInfo(id="x", input_cost_per_token="free")
+
+
+@pytest.mark.parametrize(
+    "value, expected",
+    [
+        (True, True),
+        ("true", True),
+        (" False ", False),
+        ("yes", True),
+        (None, None),
+        ("os.environ/DROP_PARAMS", "os.environ/DROP_PARAMS"),
+        ("v2:gcm:ciphertext-from-a-pre-fix-row", "v2:gcm:ciphertext-from-a-pre-fix-row"),
+    ],
+)
+def test_drop_params_coerces_flags_and_keeps_unresolved_strings(value, expected):
+    assert GenericLiteLLMParams(drop_params=value).drop_params == expected
+
+
+def test_drop_params_rejects_non_flag_non_string_values():
+    with pytest.raises(ValidationError):
+        GenericLiteLLMParams(drop_params=2)

ui/litellm-dashboard/src/lib/http/schema.d.ts

@@ -29397,7 +29397,7 @@ export interface components {
             /** Default Api Key Tpm Limit */
             default_api_key_tpm_limit?: number | null;
             /** Drop Params */
-            drop_params?: boolean | null;
+            drop_params?: boolean | string | null;
             /** Gcs Bucket Name */
             gcs_bucket_name?: string | null;
             /** Google Maps Grounding Cost Per Query */
@@ -39570,7 +39570,7 @@ export interface components {
             /** Default Api Key Tpm Limit */
             default_api_key_tpm_limit?: number | null;
             /** Drop Params */
-            drop_params?: boolean | null;
+            drop_params?: boolean | string | null;
             /** Gcs Bucket Name */
             gcs_bucket_name?: string | null;
             /** Google Maps Grounding Cost Per Query */