style: auto-format with black + isort + prettier [skip ci-format-check]

apps/worker/video_processing/dedup.py

@@ -1155,7 +1155,6 @@ def compute_structure_similarity(clips1: list[dict], clips2: list[dict]) -> floa
     # 三维度加权：数量 0.3 + 类型 0.4 + 时长 0.3
     return count_sim * 0.3 + type_sim * 0.4 + duration_sim * 0.3
 
-
     def compute_duplicate_rate(
         self,
         fingerprint: VideoFingerprint,
@@ -1205,12 +1204,16 @@ def compute_structure_similarity(clips1: list[dict], clips2: list[dict]) -> floa
 
         # Issue #P2-后端3: 加载当前视频的文案+结构数据
         from packages.adapters.sqlalchemy_impl.models import EditPlanClipModel, GeneratedVideoModel
-        
-        current_video_obj = session.query(GeneratedVideoModel).filter(GeneratedVideoModel.id == current_video_id).first() if current_video_id else None
+
+        current_video_obj = (
+            session.query(GeneratedVideoModel).filter(GeneratedVideoModel.id == current_video_id).first()
+            if current_video_id
+            else None
+        )
         current_plan_id = getattr(current_video_obj, "edit_plan_id", "") or ""
         current_clips_data = []
         current_text_content = ""
-        
+
         if current_plan_id:
             current_clips = (
                 session.query(EditPlanClipModel)
@@ -1218,10 +1221,7 @@ def compute_structure_similarity(clips1: list[dict], clips2: list[dict]) -> floa
                 .order_by(EditPlanClipModel.order)
                 .all()
             )
-            current_clips_data = [
-                {"clip_type": c.clip_type, "duration": c.duration}
-                for c in current_clips
-            ]
+            current_clips_data = [{"clip_type": c.clip_type, "duration": c.duration} for c in current_clips]
             # 拼接所有片段的文本内容
             current_text_content = " ".join(c.text_content for c in current_clips if c.text_content)
 
@@ -1299,7 +1299,7 @@ def compute_structure_similarity(clips1: list[dict], clips2: list[dict]) -> floa
             existing_plan_id = getattr(existing, "edit_plan_id", "") or ""
             existing_clips_data = []
             existing_text_content = ""
-            
+
             if existing_plan_id:
                 existing_clips = (
                     session.query(EditPlanClipModel)
@@ -1307,22 +1307,29 @@ def compute_structure_similarity(clips1: list[dict], clips2: list[dict]) -> floa
                     .order_by(EditPlanClipModel.order)
                     .all()
                 )
-                existing_clips_data = [
-                    {"clip_type": c.clip_type, "duration": c.duration}
-                    for c in existing_clips
-                ]
+                existing_clips_data = [{"clip_type": c.clip_type, "duration": c.duration} for c in existing_clips]
                 existing_text_content = " ".join(c.text_content for c in existing_clips if c.text_content)
 
             # 计算文案相似度（有文案才算）
-            text_sim = compute_text_similarity(current_text_content, existing_text_content) if (current_text_content and existing_text_content) else 0.0
+            text_sim = (
+                compute_text_similarity(current_text_content, existing_text_content)
+                if (current_text_content and existing_text_content)
+                else 0.0
+            )
 
             # 计算结构相似度（有片段才算）
-            structure_sim = compute_structure_similarity(current_clips_data, existing_clips_data) if (current_clips_data and existing_clips_data) else 0.0
+            structure_sim = (
+                compute_structure_similarity(current_clips_data, existing_clips_data)
+                if (current_clips_data and existing_clips_data)
+                else 0.0
+            )
 
             # 多维度融合：visual*0.5 + text*0.25 + structure*0.25
             # 如果文案/结构数据缺失，只用视觉维度（visual 权重提升到 1.0）
             if current_text_content and existing_text_content and current_clips_data and existing_clips_data:
-                dup_rate = (visual_sim * VISUAL_WEIGHT + text_sim * TEXT_WEIGHT + structure_sim * STRUCTURE_WEIGHT) * 100
+                dup_rate = (
+                    visual_sim * VISUAL_WEIGHT + text_sim * TEXT_WEIGHT + structure_sim * STRUCTURE_WEIGHT
+                ) * 100
             else:
                 # 降级：只有视觉维度
                 dup_rate = visual_sim * 100

tests/unit/test_dedup_enhanced.py

@@ -31,6 +31,7 @@ _mock_if_absent("packages.shared.storage")
 # Mock cv2 and numpy if not available
 try:
     import cv2 as _cv2
+
     if not isinstance(_cv2, MagicMock):
         _HAS_CV2 = True
     else:
@@ -154,11 +155,11 @@ class TestStructureSimilarity:
         """时长分布不同，时长相似度低。"""
         clips1 = [
             {"clip_type": "video", "duration": 10.0},  # 占比 80%
-            {"clip_type": "title", "duration": 2.5},    # 占比 20%
+            {"clip_type": "title", "duration": 2.5},  # 占比 20%
         ]
         clips2 = [
-            {"clip_type": "video", "duration": 2.0},   # 占比 20%
-            {"clip_type": "title", "duration": 8.0},    # 占比 80%
+            {"clip_type": "video", "duration": 2.0},  # 占比 20%
+            {"clip_type": "title", "duration": 8.0},  # 占比 80%
         ]
         sim = compute_structure_similarity(clips1, clips2)
         assert 0.7 < sim < 0.9  # 类型相同但时长分布不同，sim=0.82