Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.
Python agentaichatgptdeepseekfine-tuninggemmaimage-generationllamallmllmsopenaipythonqwenreinforcement-learningself-hostedstable-diffusiontext-to-speechttsuiunsloth
Studio tests: forget the managed provider URL setting cache around each test (#12288)
Daniel Han7b2cb03cece6
diff --git a/studio/backend/tests/conftest.py b/studio/backend/tests/conftest.pyindex 2b4d88d4896..ad9859b2fd8 100644--- a/studio/backend/tests/conftest.py+++ b/studio/backend/tests/conftest.py@@ -351,6 +351,26 @@ def _isolate_generation_state():active_generations.reset_for_tests()[email protected](autouse = True)+def _forget_the_managed_provider_url_setting():+ """Drop the managed-account private provider URL setting's cached answer around each test.++ ``managed_provider_url_settings`` holds the owner's switch for a second so every managed+ outbound request does not open SQLite. Each test points ``UNSLOTH_STUDIO_HOME`` at a fresh+ store, but the held answer outlives it: a test that turned the switch on left ``True`` cached,+ and the next test in the same xdist worker to run inside that second read it instead of its own+ store's default, so test_managed_account_cannot_list_models_from_a_loopback_provider got a 200+ with the loopback provider's models where it expected a 400.+ """+ settings = sys.modules.get("utils.managed_provider_url_settings")+ if settings is not None:+ settings.forget_cached_setting()+ yield+ settings = sys.modules.get("utils.managed_provider_url_settings")+ if settings is not None:+ settings.forget_cached_setting()++@pytest.fixture(autouse = True)def _forget_the_cached_owner_identity():"""Drop ``process_lifetime``'s cached owner identity after each test.