yeuimu
08d26710b4
- 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）
70 lines
2.5 KiB
Python
70 lines
2.5 KiB
Python
# -*- coding: utf-8 -*-
|
|
"""providers 模块测试：判定 JSON 解析、分类归一化、图片编码、usage 规整。"""
|
|
import base64
|
|
|
|
import pytest
|
|
|
|
from violation_detector.providers import CANON, _norm_usage, encode_image, norm_category, parse_judgment
|
|
|
|
GOOD = ('{"文字/图形属性": "黑色T恤印字母", "侵权/违规逻辑": "未授权使用商标", '
|
|
'"违规分类": "12. 侵权 - 除人物外的其他侵权"}')
|
|
|
|
|
|
def test_parse_plain_json():
|
|
r = parse_judgment(GOOD)
|
|
assert r is not None
|
|
assert r[0] == "黑色T恤印字母"
|
|
assert r[1] == "未授权使用商标"
|
|
assert r[2] == "12. 侵权 - 除人物外的其他侵权"
|
|
|
|
|
|
def test_parse_fenced_json():
|
|
r = parse_judgment("```json\n" + GOOD + "\n```")
|
|
assert r is not None and r[2].startswith("12.")
|
|
|
|
|
|
def test_parse_json_with_surrounding_text():
|
|
r = parse_judgment("分析如下：\n" + GOOD + "\n以上。")
|
|
assert r is not None
|
|
|
|
|
|
def test_parse_invalid():
|
|
assert parse_judgment("") is None
|
|
assert parse_judgment("这不是JSON") is None
|
|
assert parse_judgment('{"文字/图形属性": "x"}') is None # 缺字段
|
|
assert parse_judgment('{"a": [1,2]') is None # 截断
|
|
|
|
|
|
def test_norm_category_variants():
|
|
# 不同写法归一到标准 17 类名称
|
|
assert norm_category("12. 侵权 - 人物外") == CANON[12]
|
|
assert norm_category("14脏话、侮辱性") == CANON[14]
|
|
assert norm_category("5. 负向敏感") == CANON[5]
|
|
# 非编号类原样保留
|
|
assert norm_category("无违规") == "无违规"
|
|
assert norm_category("违规不明") == "违规不明"
|
|
# 未知编号保留原文
|
|
assert norm_category("99. 未来分类") == "99. 未来分类"
|
|
|
|
|
|
def test_encode_image_roundtrip(tmp_path):
|
|
img = tmp_path / "a.jpg"
|
|
img.write_bytes(b"\xff\xd8\xff\xe0fake")
|
|
url = encode_image(str(img))
|
|
assert url.startswith("data:image/jpeg;base64,")
|
|
assert base64.b64decode(url.split(",", 1)[1]) == b"\xff\xd8\xff\xe0fake"
|
|
|
|
|
|
def test_encode_image_unsupported(tmp_path):
|
|
p = tmp_path / "a.txt"
|
|
p.write_text("x", encoding="utf-8")
|
|
with pytest.raises(ValueError):
|
|
encode_image(str(p))
|
|
|
|
|
|
def test_norm_usage_handles_missing_details():
|
|
assert _norm_usage({"prompt_tokens": 100, "completion_tokens": 50}) == {
|
|
"prompt": 100, "completion": 50, "cached": 0}
|
|
assert _norm_usage({}) == {"prompt": None, "completion": None, "cached": 0}
|
|
assert _norm_usage({"prompt_tokens": 100, "prompt_tokens_details": {"cached_tokens": 80}})["cached"] == 80
|