diff --git a/litellm/__init__.py b/litellm/__init__.py
index f949046e4d3..7277f4e7d5e 100644
--- a/litellm/__init__.py
+++ b/litellm/__init__.py
@@ -16,7 +16,8 @@ import os
 # Load .env before any other litellm imports so env vars (e.g. LITELLM_UI_SESSION_DURATION) are available
 import dotenv as _dotenv
 
-_dotenv.load_dotenv()
+if os.getenv("LITELLM_MODE", "DEV") == "DEV":
+    _dotenv.load_dotenv()
 
 from typing import (
     Callable,
diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py
index 607306f3806..2ed40213fb7 100644
--- a/litellm/proxy/proxy_server.py
+++ b/litellm/proxy/proxy_server.py
@@ -50,6 +50,7 @@ from litellm.constants import (
     LITELLM_EMBEDDING_PROVIDERS_SUPPORTING_INPUT_ARRAY_OF_TOKENS,
     LITELLM_SETTINGS_SAFE_DB_OVERRIDES,
     LITELLM_UI_ALLOW_HEADERS,
+    LITELLM_UI_SESSION_DURATION,
 )
 from litellm.litellm_core_utils.litellm_logging import (
     _init_custom_logger_compatible_class,
@@ -10767,7 +10768,7 @@ async def onboarding(invite_link: str, request: Request):
         request_type="key",
         **{
             "user_role": user_obj.user_role,
-            "duration": "24hr",
+            "duration": LITELLM_UI_SESSION_DURATION,
             "key_max_budget": litellm.max_ui_session_budget,
             "models": [],
             "aliases": {},