fix: enrich cleanup from active queue context

tests/test_queue_cleanup.py

@@ -65,6 +65,13 @@ class FakeQueueClient:
         self.records = [record for record in self.records if int(record["taskId"]) not in ids]
         return {"code": 0, "data": None}
 
+    def list_framework_stats(self, task_type: str, target_gpu: str) -> list[dict[str, Any]]:
+        del task_type, target_gpu
+        return [
+            {"framework": "vllm", "modelCount": 100, "successCount": 60},
+            {"framework": "transformers", "modelCount": 100, "successCount": 50},
+        ]
+
     def count_active_tasks(self, **_kwargs: Any) -> int:
         return len(self.records)
 
@@ -173,6 +180,107 @@ class QueueCleanupTests(unittest.TestCase):
         self.assertEqual(1, skipped["runningMatchedProtected"])
         self.assertEqual(1, skipped["noMatchingBlock"])
 
+    def test_architecture_cleanup_uses_current_queue_task_type_when_every_framework_is_blocked(self) -> None:
+        vllm_key, vllm_block = self.architecture_block(framework="vllm")
+        transformers_key, transformers_block = self.architecture_block(framework="transformers")
+        tasks = [
+            OwnedTask(
+                0,
+                1,
+                "owner/model",
+                "Iluvatar_bi-100",
+                "waiting",
+                task_type="text-generation",
+            )
+        ]
+
+        selected, skipped = find_architecture_incompatible_tasks(
+            tasks,
+            architecture_blocks={
+                vllm_key: vllm_block,
+                transformers_key: transformers_block,
+            },
+            task_contexts={},
+            model_configs={
+                "owner/model": {"architectures": ["Qwen2ForCausalLM"]},
+            },
+            framework_catalog={
+                ("iluvatar_bi-100", "text-generation"): {"vllm", "transformers"},
+            },
+        )
+
+        self.assertEqual([1], [item["taskId"] for item in selected])
+        self.assertEqual("all_live_frameworks", selected[0]["architectureMatchScope"])
+        self.assertEqual(["transformers", "vllm"], selected[0]["evaluatedFrameworks"])
+        self.assertEqual(1, skipped["frameworkContextUnknown"])
+
+    def test_architecture_cleanup_does_not_guess_when_only_some_frameworks_are_blocked(self) -> None:
+        key, block = self.architecture_block(framework="vllm")
+        tasks = [
+            OwnedTask(
+                0,
+                1,
+                "owner/model",
+                "Iluvatar_bi-100",
+                "waiting",
+                task_type="text-generation",
+            )
+        ]
+
+        selected, skipped = find_architecture_incompatible_tasks(
+            tasks,
+            architecture_blocks={key: block},
+            task_contexts={},
+            model_configs={
+                "owner/model": {"architectures": ["Qwen2ForCausalLM"]},
+            },
+            framework_catalog={
+                ("iluvatar_bi-100", "text-generation"): {"vllm", "transformers"},
+            },
+        )
+
+        self.assertEqual([], selected)
+        self.assertEqual(1, skipped["partiallyBlockedFrameworkSet"])
+
+    def test_queue_cleanup_enriches_active_api_record_and_stops_only_all_framework_block(self) -> None:
+        vllm_key, vllm_block = self.architecture_block(framework="vllm")
+        transformers_key, transformers_block = self.architecture_block(framework="transformers")
+        client = FakeQueueClient(
+            [
+                {
+                    "taskId": 1,
+                    "modelId": "owner/model",
+                    "gpuType": "Iluvatar_bi-100",
+                    "status": "waiting",
+                    "modelTaskLevelId": 23,
+                    "modelTaskLevel": "文本生成",
+                }
+            ]
+        )
+        pool = ModelHubClientPool([client], active_task_cap=100)  # type: ignore[list-item]
+
+        summary = cleanup_certain_oom_tasks(
+            pool,
+            FakeDiscovery(
+                {"owner/model": 1 * GIB},
+                configs={"owner/model": {"architectures": ["Qwen2ForCausalLM"]}},
+            ),  # type: ignore[arg-type]
+            architecture_compatibility_blocks={
+                vllm_key: vllm_block,
+                transformers_key: transformers_block,
+            },
+            task_compatibility_contexts={},
+            log=lambda _message: None,
+        )
+
+        self.assertEqual(1, summary["architectureIncompatibleCount"])
+        self.assertEqual(1, summary["cancelledCount"])
+        self.assertEqual(
+            ["transformers", "vllm"],
+            summary["architectureFrameworkCatalog"]["iluvatar_bi-100|text-generation"],
+        )
+        self.assertEqual([[1]], client.stopped)
+
     def test_queue_cleanup_fetches_config_and_stops_known_incompatible_waiting_task(self) -> None:
         key, block = self.architecture_block()
         client = FakeQueueClient(