test(tree): 建树模块端到端集成测试

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

tests/integration/test_tree_build_e2e.py

@@ -0,0 +1,172 @@
+"""建树模块端到端集成测试。
+
+验证各模块协作：
+  构造最小树 → verify → subtitle 注入 → TreeEnvironment 查询 → 序列化 roundtrip
+"""
+
+from __future__ import annotations
+
+import numpy as np
+import pytest
+
+from app.tree.index import (
+    IndexMeta, TreeIndex, L1Node, L1Card,
+    L2Node, L2Card, L3Node, L3Card,
+)
+from app.tree.verify import verify_tree
+from app.tree.subtitle import SRTEntry, assign_subtitles_voronoi
+from app.tree.environment import TreeEnvironment
+
+
+class TestTreeModuleE2E:
+    def test_verify_subtitle_environment_pipeline(self, tmp_path):
+        """完整流程：构造树 → verify（删除幻觉实体）→ subtitle 注入 → environment 查询 → 序列化 roundtrip。"""
+        # 构造一棵树，L2 有混合实体（有出处/无出处）
+        l3_0 = L3Node(
+            id="vid_L1_000_L2_000_L3_000",
+            card=L3Card(
+                frame_summary="运动员在跑步冲刺",
+                visible_entities=["运动员", "跑道"],
+                ongoing_actions=["跑步"],
+                visible_text=["Nike", "2024"],
+                spatial_layout="居中构图",
+                visual_attributes={"lighting": "明亮", "camera_angle": "侧面"},
+            ),
+            timestamp=2.0,
+            frame_path="frames/L1_000_L2_000_L3_000.jpg",
+        )
+        l3_1 = L3Node(
+            id="vid_L1_000_L2_000_L3_001",
+            card=L3Card(
+                frame_summary="观众在看台上欢呼",
+                visible_entities=["观众", "看台"],
+                ongoing_actions=["欢呼"],
+                visible_text=["Stadium"],
+                spatial_layout="广角",
+                visual_attributes={},
+            ),
+            timestamp=6.0,
+            frame_path="frames/L1_000_L2_000_L3_001.jpg",
+        )
+        l2 = L2Node(
+            id="vid_L1_000_L2_000",
+            card=L2Card(
+                event_description="百米决赛片段",
+                entities=["运动员", "裁判", "幻觉实体XYZ"],
+                actions=["跑步", "欢呼"],
+                action_subjects=["运动员", "观众"],
+                visible_text=["Nike", "不存在的文字ABC"],
+                spatial_relations="运动员在跑道中央",
+                state_changes=None,
+            ),
+            time_range=(0.0, 10.0),
+            children=[l3_0, l3_1],
+        )
+        l1 = L1Node(
+            id="vid_L1_000",
+            card=L1Card(
+                scene_summary="百米短跑决赛",
+                main_setting="体育场",
+                key_entities=["运动员", "不存在的人物"],
+                main_actions=["比赛"],
+                topic_keywords=["体育", "短跑"],
+                visible_text=["Nike", "Ghost文字"],
+                temporal_flow="从起跑到冲刺",
+            ),
+            time_range=(0.0, 10.0),
+            children=[l2],
+        )
+        index = TreeIndex(
+            metadata=IndexMeta(source_path="/test/video.mp4", modality="video"),
+            roots=[l1],
+        )
+
+        # Step 1: verify — 删除无出处的实体和 visible_text
+        stats = verify_tree(index)
+        assert "幻觉实体XYZ" not in index.roots[0].children[0].card.entities
+        assert "不存在的文字ABC" not in index.roots[0].children[0].card.visible_text
+        assert "Ghost文字" not in index.roots[0].card.visible_text
+        assert "不存在的人物" not in index.roots[0].card.key_entities
+        # 有出处的保留
+        assert "运动员" in index.roots[0].children[0].card.entities
+        assert "Nike" in index.roots[0].children[0].card.visible_text
+
+        # Step 2: subtitle 注入
+        srt_entries = [
+            SRTEntry(start=1.0, end=3.0, text="And the runner sprints ahead!"),
+            SRTEntry(start=5.0, end=7.0, text="The crowd goes wild!"),
+        ]
+        assign_subtitles_voronoi(index, srt_entries)
+        assert l3_0.subtitle is not None
+        assert "sprints" in l3_0.subtitle
+        assert l3_1.subtitle is not None
+        assert "crowd" in l3_1.subtitle
+
+        # Step 3: TreeEnvironment 查询
+        env = TreeEnvironment(index)
+
+        # view_node L3
+        l3_view = env.view_node("vid_L1_000_L2_000_L3_000")
+        assert "运动员在跑步冲刺" in l3_view
+
+        # view_node L2 (should list children)
+        l2_view = env.view_node("vid_L1_000_L2_000")
+        assert "百米决赛片段" in l2_view
+        assert "vid_L1_000_L2_000_L3_000" in l2_view
+
+        # view_node with anchor
+        anchored = env.view_node("vid_L1_000_L2_000_L3_000", anchor=True)
+        assert "[c" in anchored
+
+        # get_subtitle
+        assert "sprints" in env.get_subtitle("vid_L1_000_L2_000_L3_000")
+
+        # search_similar (with embedding)
+        def fake_embed(texts):
+            if isinstance(texts, str):
+                texts = [texts]
+            rng = np.random.RandomState(42)
+            return rng.randn(len(texts), 4).astype(np.float32)
+
+        index.embed_all(fake_embed, "test-model", 4)
+        results = env.search_similar("运动员跑步", top_k=3, embed_fn=fake_embed)
+        assert len(results) > 0
+
+        # Step 4: 序列化 roundtrip
+        path = tmp_path / "tree.json"
+        index.save_json(str(path))
+        loaded = TreeIndex.load_json(str(path))
+
+        assert len(loaded.roots) == 1
+        assert loaded.roots[0].card.scene_summary == "百米短跑决赛"
+        assert loaded.roots[0].children[0].children[0].subtitle is not None
+        assert "sprints" in loaded.roots[0].children[0].children[0].subtitle
+        # verify 的修改也被保留
+        assert "幻觉实体XYZ" not in loaded.roots[0].children[0].card.entities
+
+    def test_repair_detector_on_broken_tree(self):
+        """修复检测器能识别空卡片节点。"""
+        from app.tree.repair.detector import detect_issues
+
+        l3 = L3Node(
+            id="vid_L1_000_L2_000_L3_000",
+            card=L3Card("", [], [], [], "", {}),  # empty frame_summary
+            timestamp=1.0,
+        )
+        l2 = L2Node(
+            id="vid_L1_000_L2_000",
+            card=L2Card("事件", [], [], [], [], "", None),
+            time_range=(0.0, 10.0),
+            children=[l3],
+        )
+        l1 = L1Node(
+            id="vid_L1_000",
+            card=L1Card("场景", "", [], [], [], [], ""),
+            time_range=(0.0, 10.0),
+            children=[l2],
+        )
+        index = TreeIndex(metadata=IndexMeta("/t.mp4", "video"), roots=[l1])
+
+        issues = detect_issues(index)
+        assert len(issues) >= 1
+        assert any(i.issue_type == "empty_field" for i in issues)