Files
AgentLab/backend/tests/test_report_export_masked_metrics.py
T

86 lines
4.2 KiB
Python

"""报告导出成 HTML 时,取自遮罩格的口径卡指标不随文件出去(报告导出与可验证报告的遮罩改造合并时补上)。
报告导出照文档原样渲染正文、按证据面板的拼法嵌出处链。现在的报告撰写节点把这类指标写成「已遮罩」,导出的文件里本来就
没有原值;升级前写的老文档(或写完之后数据源才设的遮罩)证据目录里还带着原值,正文引用了的话正文里也是——可验证报告把
这样的文档整份只给哈希,报告导出同样会拿给系统外的人,按同一个判断把引用了它们的片段写成「已遮罩」,正文、出处链里都
不出现原值;普通指标照常。
真跑的部分用 tests/test_report_masked_metrics.py 的虚构会员库(号码是编的),写作者是 MockChatModel 照剧本写。
"""
from __future__ import annotations
import json
import re
from tests.test_report_masked_metrics import ( # noqa: F401 - 夹具按名字引入
CITES,
CLEAN,
client,
engine_up,
finish,
graph,
leaks,
script,
vip,
)
async def _export(client, run_id: str) -> tuple[str, dict]:
res = await client.get(f"/api/runs/{run_id}/report.html")
assert res.status_code == 200, res.text
html = res.text
block = re.search(r'<script type="application/json" id="report-data">(.*?)</script>', html, re.S)
assert block, "页面里有数据块"
return html, json.loads(block.group(1))
def _metric_step(data: dict, metric: str) -> dict:
return next(step for chain in data["chains"].values() for step in chain
if step.get("step") == "metric" and step.get("metric") == metric)
async def test_a_new_doc_exports_without_the_value(client, monkeypatch, vip):
"""现在的报告撰写节点写的文档(写作者坚持引用了 [[m:phone_raw]],标注照常产出):正文、出处链里都是「已遮罩」。"""
script(monkeypatch, CITES)
row = await finish(graph(on_violation="flag"))
assert row.status == "succeeded", row.error
html, data = await _export(client, row.id)
assert "手机号 </span>" in html and ">已遮罩</button>" in html
step = _metric_step(data, "phone_raw")
assert (step["value"], step["rendered"], step["render_ok"]) == ("已遮罩", "已遮罩", True), step
assert leaks(html) == []
async def test_an_old_doc_exports_the_masked_metric_as_masked(client, monkeypatch, vip):
"""升级前的报告撰写节点写的文档(这里把写作前认遮罩的那一步换成什么都不认):目录、正文里写着号码。导出时引用它的那一处
写「已遮罩」,出处链的值、渲染也是,渲染一致;同一句里的普通指标(客户编号)照常;文件里一个号码都没有。"""
import app.api.evidence as evidence_api
async def nothing(*_args, **_kwargs):
return []
real = evidence_api.withheld_metrics
monkeypatch.setattr(evidence_api, "withheld_metrics", nothing)
script(monkeypatch, CITES)
row = await finish(graph(on_violation="flag"))
assert row.status == "succeeded", row.error
monkeypatch.setattr(evidence_api, "withheld_metrics", real)
html, data = await _export(client, row.id)
assert ">已遮罩</button>" in html and "title=\"数字「已遮罩」:点开查看出处\"" in html
step = _metric_step(data, "phone_raw")
assert (step["value"], step["rendered"], step["render_ok"]) == ("已遮罩", "已遮罩", True), step
first = _metric_step(data, "first_id")
assert (first["value"], first["rendered"], first["render_ok"]) == (1, "1", True), first
assert leaks(html) == []
async def test_ordinary_metrics_export_as_before(client, monkeypatch, vip):
"""只引用普通指标的报告:导出照旧,客户数、编号照写,出处链照常;文件里没有号码(会员查询的行里号码列写「已遮罩」)。"""
script(monkeypatch, CLEAN)
row = await finish(graph(on_violation="flag"))
assert row.status == "succeeded", row.error
html, data = await _export(client, row.id)
assert ">3位</button>" in html and ">1</button>" in html
assert _metric_step(data, "n_cust")["value"] == 3
assert leaks(html) == []