init: 商品图合规检测工具（豆包初筛 + DeepSeek 复检级联）

- src 标准布局：config/providers/pipeline/report + CLI/Tk GUI 双入口 - 级联省钱：豆包全量初筛，仅无违规/违规不明图进 DeepSeek 复检（含两票复核） - 输出：时间戳目录 + 分类文件夹图片归档 + Excel 报表 - 30 个单元测试（tests/，测试图片不入库）
2026-09-02 15:28:51 +08:00
commit 08d26710b4
20 changed files with 2513 additions and 0 deletions
@@ -0,0 +1,73 @@
+# -*- coding: utf-8 -*-
+"""report 模块测试：目录名清洗、输出组织（时间戳+分类归档）、Excel 报表。"""
+from openpyxl import load_workbook
+
+from violation_detector.report import build_report, organize_output, sanitize_dirname
+
+
+def ROW(cat, fname="img.jpg", channel="豆包初筛"):
+    return {"file": fname, "category": cat, "channel": channel, "attr": "属性",
+            "logic": "逻辑", "status": "ok", "vote_note": "", "remark": ""}
+
+
+def test_sanitize_dirname():
+    assert sanitize_dirname("12. 侵权 - 其他") == "12. 侵权 - 其他"
+    bad = 'a<b>c:"d/e\\f|g?h*i'
+    clean = sanitize_dirname(bad)
+    for ch in '<>:"/\\|?*':
+        assert ch not in clean
+    assert clean == "a_b_c__d_e_f_g_h_i"
+    assert sanitize_dirname("  . ") == "未分类"
+    assert len(sanitize_dirname("长" * 100)) <= 80
+
+
+def test_organize_output_copies_images_by_category(tmp_path):
+    src = tmp_path / "src"
+    src.mkdir()
+    for name in ("a.jpg", "b.jpg", "c.jpg"):
+        (src / name).write_bytes(b"img")
+    rows = [ROW("12. 侵权 - 除人物外的其他侵权", "a.jpg"),
+            ROW("无违规", "b.jpg"),
+            ROW("12. 侵权 - 除人物外的其他侵权", "c.jpg")]
+    ts_dir = organize_output(rows, str(src), str(tmp_path / "out"))
+    assert ts_dir.name.startswith("检测结果_")
+    assert (ts_dir / "12. 侵权 - 除人物外的其他侵权" / "a.jpg").read_bytes() == b"img"
+    assert (ts_dir / "12. 侵权 - 除人物外的其他侵权" / "c.jpg").exists()
+    assert (ts_dir / "无违规" / "b.jpg").exists()
+    # 源文件保留（复制而非移动）
+    assert (src / "a.jpg").exists()
+
+
+def test_organize_output_missing_image_no_crash(tmp_path):
+    rows = [ROW("无违规", "不存在.jpg")]
+    ts_dir = organize_output(rows, str(tmp_path), str(tmp_path / "out"))
+    assert (ts_dir / "无违规").is_dir()
+
+
+def test_build_report_structure(tmp_path):
+    rows = [ROW("12. 侵权 - 除人物外的其他侵权", "a.jpg"),
+            ROW("无违规", "b.jpg", channel="DeepSeek复检"),
+            ROW("无违规", "b.jpg")]
+    rows[2]["vote_note"] = "票型2:1（无违规 / 无违规 / 12. 侵权），建议人工复核"
+    out = build_report(rows, str(tmp_path), {"mode": "cascade", "ds_calls": 3,
+                                             "ds_peak_cost": 0.1, "ds_idle_cost": 0.05})
+    assert out.endswith(".xlsx")
+
+    wb = load_workbook(out)
+    ws = wb["检测结果"]
+    assert ws["B2"].value.startswith("商品图合规检测工具")
+    headers = [ws.cell(row=4, column=c).value for c in range(2, 10)]
+    assert headers == ["序号", "文件名", "违规分类", "判定通道", "文字/图形属性",
+                       "侵权/违规逻辑", "票型/复核", "备注"]
+    assert ws.cell(row=5, column=2).value == 1
+    assert ws.cell(row=6, column=5).value == "DeepSeek复检"
+    assert ws.cell(row=5, column=5).value == "豆包初筛"
+    assert "人工复核" in ws.cell(row=7, column=8).value
+
+    ws2 = wb["统计汇总"]
+    cats = [ws2.cell(row=r, column=2).value for r in (5, 6)]
+    assert cats[0].startswith("12.") and cats[1] == "无违规"
+    assert ws2.cell(row=5, column=3).value == 1
+    assert ws2.cell(row=6, column=3).value == 2
+    # 合计行
+    assert ws2.cell(row=7, column=3).value == 3