fix(tests): move llama stack client init back to fixture (#3071)

See inline comments

tests/integration/conftest.py

@@ -9,15 +9,12 @@ import os
 import platform
 import textwrap
 import time
-import warnings
 
 import pytest
 from dotenv import load_dotenv
 
 from llama_stack.log import get_logger
 
-from .fixtures.common import instantiate_llama_stack_client
-
 logger = get_logger(__name__, category="tests")
 
 
@@ -34,20 +31,6 @@ def pytest_sessionstart(session):
     # stop macOS from complaining about duplicate OpenMP libraries
     os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"
 
-    # pull client instantiation to session start so all the complex logs during initialization
-    # don't clobber the test one-liner outputs
-    print("instantiating llama_stack_client")
-    start_time = time.time()
-    with warnings.catch_warnings():
-        warnings.filterwarnings("ignore", category=DeprecationWarning)
-
-        try:
-            session._llama_stack_client = instantiate_llama_stack_client(session)
-        except Exception as e:
-            logger.error(f"Error instantiating llama_stack_client: {e}")
-            session._llama_stack_client = None
-    print(f"llama_stack_client instantiated in {time.time() - start_time:.3f}s")
-
 
 def pytest_runtest_teardown(item):
     # Check if the test actually ran and passed or failed, but was not skipped or an expected failure (xfail)

tests/integration/fixtures/common.py

@@ -178,8 +178,13 @@ def skip_if_no_model(request):
 
 @pytest.fixture(scope="session")
 def llama_stack_client(request):
-    client = request.session._llama_stack_client
-    assert client is not None, "llama_stack_client not found in session cache"
+    # ideally, we could do this in session start given all the complex logs during initialization
+    # don't clobber the test one-liner outputs. however, this also means all tests in a sub-directory
+    # would be forced to use llama_stack_client, which is not what we want.
+    print("\ninstantiating llama_stack_client")
+    start_time = time.time()
+    client = instantiate_llama_stack_client(request.session)
+    print(f"llama_stack_client instantiated in {time.time() - start_time:.3f}s")
     return client