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: pin which rejection site the confirm-stream tests assert (#10284)
Daniel Han0bb033e3877c
diff --git a/studio/backend/tests/test_openai_tool_passthrough.py b/studio/backend/tests/test_openai_tool_passthrough.pyindex fe7d0a0292c..382d71d022c 100644--- a/studio/backend/tests/test_openai_tool_passthrough.py+++ b/studio/backend/tests/test_openai_tool_passthrough.py@@ -2341,7 +2341,11 @@ class TestChatCompletionRequestToolFields:)self._assert_unsupported_param(resp, param)- def test_confirm_tool_calls_requires_streaming_for_safetensors_tools(self, monkeypatch):+ # Same two rejection sites, and same reason to pin one, as the GGUF case below.+ @pytest.mark.parametrize("validate_before_switch", [False, True])+ def test_confirm_tool_calls_requires_streaming_for_safetensors_tools(+ self, monkeypatch, validate_before_switch+ ):import routes.inference as inference_routeclass _NoGGUFBackend:@@ -2363,6 +2367,9 @@ class TestChatCompletionRequestToolFields:"_detect_safetensors_features",lambda backend, chat_template, tools = None: {"supports_tools": True},)+ monkeypatch.setattr(+ inference_route, "_should_validate_before_switch", lambda: validate_before_switch+ )monitor = ApiMonitor(max_entries = 3)monkeypatch.setattr(inference_route, "api_monitor", monitor)client = self._v1_client(monkeypatch, _NoGGUFBackend(), _InferenceBackend())@@ -2381,9 +2388,13 @@ class TestChatCompletionRequestToolFields:body = resp.json()assert body["error"]["param"] == "confirm_tool_calls"assert "requires stream=true" in body["error"]["message"]- [entry] = monitor.snapshot()- assert entry["status"] == "error"- assert "confirm_tool_calls requires stream=true" in entry["error"]+ if validate_before_switch:+ # Early site rejects before the monitor row is opened.+ assert monitor.snapshot() == []+ else:+ [entry] = monitor.snapshot()+ assert entry["status"] == "error"+ assert "confirm_tool_calls requires stream=true" in entry["error"]assert monitor.active_count() == 0def test_multiturn_tool_loop_messages(self):@@ -4195,7 +4206,14 @@ class TestGgufVisionToolRouting:assert exc.value.status_code == 400assert exc.value.detail["error"]["param"] == "tool_choice"- def test_confirm_tool_calls_requires_streaming_for_gguf_tools(self, monkeypatch):+ # Two sites reject this shape and `_should_validate_before_switch()` picks+ # which; the early one runs before the api_monitor row is opened. It reads+ # process-wide state, so unpinned this asserted whichever site the xdist+ # worker happened to leave reachable. Pinned, both sites are covered.+ @pytest.mark.parametrize("validate_before_switch", [False, True])+ def test_confirm_tool_calls_requires_streaming_for_gguf_tools(+ self, monkeypatch, validate_before_switch+ ):import routes.inference as inf_moddef _plain(**kwargs):@@ -4214,6 +4232,9 @@ class TestGgufVisionToolRouting:generate_chat_completion_with_tools = _tools,)monkeypatch.setattr(inf_mod, "get_llama_cpp_backend", lambda: backend)+ monkeypatch.setattr(+ inf_mod, "_should_validate_before_switch", lambda: validate_before_switch+ )monitor = ApiMonitor(max_entries = 3)monkeypatch.setattr(inf_mod, "api_monitor", monitor)@@ -4236,9 +4257,15 @@ class TestGgufVisionToolRouting:)assert exc.value.status_code == 400assert "requires stream=true" in exc.value.detail["error"]["message"]- [entry] = monitor.snapshot()- assert entry["status"] == "error"- assert "confirm_tool_calls requires stream=true" in entry["error"]+ if validate_before_switch:+ # Early site rejects before the monitor row is opened, so there is no+ # row to fail; opening one here would be a behaviour change, not a fix.+ assert monitor.snapshot() == []+ else:+ [entry] = monitor.snapshot()+ assert entry["status"] == "error"+ assert "confirm_tool_calls requires stream=true" in entry["error"]+ # Neither site may leave a row running.assert monitor.active_count() == 0def test_standard_gguf_stream_splits_reasoning_content(self, monkeypatch):