From cd60e4d46abb37e4a3733545ab4feb7987f4b80a Mon Sep 17 00:00:00 2001
From: Ashwin Bharambe <ashwin.bharambe@gmail.com>
Date: Wed, 3 Sep 2025 15:38:39 -0700
Subject: [PATCH] fix(tests): set inference mode to be replay by default

---
 tests/integration/conftest.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/tests/integration/conftest.py b/tests/integration/conftest.py
index 234d762ce..fd9a54d04 100644
--- a/tests/integration/conftest.py
+++ b/tests/integration/conftest.py
@@ -30,6 +30,8 @@ def pytest_runtest_makereport(item, call):
 def pytest_sessionstart(session):
     # stop macOS from complaining about duplicate OpenMP libraries
     os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"
+    if "LLAMA_STACK_TEST_INFERENCE_MODE" not in os.environ:
+        os.environ["LLAMA_STACK_TEST_INFERENCE_MODE"] = "replay"
 
 
 def pytest_runtest_teardown(item):