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
Nudge-retry cancel test: bound the hang, not the runner's speed (#13750)
Daniel Han01a130beeb12
diff --git a/studio/backend/tests/test_passthrough_healing.py b/studio/backend/tests/test_passthrough_healing.pyindex cd55be25fab..00e29ac3dee 100644--- a/studio/backend/tests/test_passthrough_healing.py+++ b/studio/backend/tests/test_passthrough_healing.py@@ -931,11 +931,15 @@ class TestNudgeRetryOpenai:cancel_event = cancel_event,))- await asyncio.wait_for(hanging.started.wait(), 0.2)+ # Both waits only bound a hang, so they are generous: reaching the retry is a whole+ # upstream round trip plus the nudge decision, 30-60 ms idle but past 1 s on a+ # saturated CI runner, which a 0.2 s cap turned into a TimeoutError. An absorbed+ # cancel still fails: the task then waits on the hanging client and never raises.+ await asyncio.wait_for(hanging.started.wait(), 10)cancel_event.set()with pytest.raises(inf_mod._NonStreamingRequestCancelled):- await asyncio.wait_for(task, 0.5)+ await asyncio.wait_for(task, 10)assert hanging.closed.is_set()