diff --git a/backend/app/engine/labels.py b/backend/app/engine/labels.py index 721bb032..1ba796a7 100644 --- a/backend/app/engine/labels.py +++ b/backend/app/engine/labels.py @@ -66,6 +66,24 @@ FIELD_LABEL: dict[str, str] = { "defaults": "工作流默认设置", } +#: 双盲复算节点各配置项的叫法(lib/terms.ts 的 DUAL_CHECK_TEXT.field,check-verify 逐项比对)。模型两项照结论句裁判 +#: 「裁判模型 · 接入」「裁判模型」的写法 +DUAL_CHECK_FIELD_LABEL: dict[str, str] = { + "card": "口径卡", "metrics": "复算的指标", "definitions": "口径说明", "provider": "盲算模型 · 接入", + "model": "盲算模型", "tolerance": "允许的相对误差", "on_reject": "未采用原流程时", "enabled": "启用双盲复算", +} +#: 「未采用原流程时」两个取值的短说法:检查器下拉的选项以它开头(DUAL_CHECK_TEXT.onReject),后面可以带一句用途 +DUAL_CHECK_ON_REJECT_LABEL: dict[str, str] = {"degrade": "计入缺口、出具降档", "fail": "判为失败"} +#: 反方审稿节点各配置项的叫法(lib/terms.ts 的 CHALLENGE_TEXT.field,check-challenge 逐项比对) +CHALLENGE_LABEL: dict[str, str] = { + "report": "要审的报告", "max_claims": "最多审几句", "max_queries": "每句最多查询次数", + "timeout_s": "每句时长上限(秒)", "provider": "反方模型 · 接入", "model": "反方模型", + "on_refuted": "结论被推翻时", "enabled": "启用", +} +#: 「结论被推翻时」三个取值的叫法:和检查器下拉(CHALLENGE_TEXT.onRefuted)一致,只去掉「(默认)」。报错让人去改 +#: 某一项时,人在下拉里找得到这个名字 +CHALLENGE_ON_REFUTED_LABEL: dict[str, str] = {"degrade": "出具降档", "flag": "只标注", "fail": "报错并停止运行"} + _FIELD_BY_TYPE: dict[str, dict[str, str]] = { "input": {"fields": "输入字段"}, "agent": {"prompt": "任务"}, @@ -81,6 +99,8 @@ _FIELD_BY_TYPE: dict[str, dict[str, str]] = { "stop_on_reject": "驳回即终止运行"}, "validate": {"source": "待校验内容", "fail_fast": "校验失败即中断"}, "merge": {"inputs": "输入", "sql": "合并 SQL"}, + "dual_check": DUAL_CHECK_FIELD_LABEL, + "challenge": CHALLENGE_LABEL, } #: 出具契约里的键(canvas/issues.ts 的 CONTRACT_KEYS、SUB_KEYS) @@ -114,6 +134,8 @@ _OPTION_LABEL: dict[str, dict[str, str]] = { "incomparable": "标注与历史不可比"}, "isolation": {"strict": "严格:microVM", "fast": "快速:系统沙箱"}, "thinking": {"summarized": "开启并显示摘要", "adaptive": "开启但不显示", "off": "关闭"}, + "on_reject": DUAL_CHECK_ON_REJECT_LABEL, + "on_refuted": CHALLENGE_ON_REFUTED_LABEL, } #: 让人在几个取值之间选时的完整选项文字(发布前修复的候选),和检查器下拉一致,只去掉「(默认)」 CHOICE_LABEL: dict[str, dict[str, str]] = { diff --git a/backend/app/engine/nodes/challenge.py b/backend/app/engine/nodes/challenge.py index f6195010..b856f34e 100644 --- a/backend/app/engine/nodes/challenge.py +++ b/backend/app/engine/nodes/challenge.py @@ -55,8 +55,8 @@ from app.engine.evidence import ENTITY_KINDS, iter_units, ledger_enabled, next_e from app.engine.judge import _usd from app.engine.merge_query import MERGE_SOURCE from app.engine.replay import _in_graph -from app.engine.schema import CHALLENGE_LABEL, CHALLENGE_ON_REFUTED_LABEL, GraphSpec, NodeType, challenge_problems, \ - challenge_report +from app.engine.labels import CHALLENGE_LABEL, CHALLENGE_ON_REFUTED_LABEL +from app.engine.schema import GraphSpec, NodeType, challenge_problems, challenge_report from app.engine.state import GraphState from app.engine.verify_agent import RATIONALE, VerifyOutcome, run_verify_agent from app.tools.datasource import run_versions diff --git a/backend/app/engine/nodes/dual_check.py b/backend/app/engine/nodes/dual_check.py index 8d30c94c..9b54b752 100644 --- a/backend/app/engine/nodes/dual_check.py +++ b/backend/app/engine/nodes/dual_check.py @@ -62,7 +62,8 @@ from app.engine.evidence import build_catalog, is_number, next_exec from app.engine.expressions import ExpressionError, _cell_call, _chain_root, cell_value, column_kind, numeric_text, \ parse_expression from app.engine.replay import _in_graph, ask, once -from app.engine.schema import DUAL_CHECK_FIELD_LABEL, DUAL_CHECK_ON_REJECT, NodeType, dual_check_card +from app.engine.labels import DUAL_CHECK_FIELD_LABEL, DUAL_CHECK_ON_REJECT_LABEL +from app.engine.schema import DUAL_CHECK_ON_REJECT, NodeType, dual_check_card from app.engine.state import GraphState #: 节点产出的状态:跳过、全部一致、核成了一部分(其余无法复算)、一个都没核成、人已裁决 @@ -208,7 +209,7 @@ async def run_dual_check(state: GraphState, ctx: NodeContext) -> dict[str, Any]: # 目录照样写了:审核人的选择是对数据的判断,和这次运行停不停无关 names = "、".join(f"「{r['name']}」" for r in rejected[:5]) raise NodeError(ctx.node.id, f"双盲复算不一致,审核人没有采用原流程的结果(指标{names})。" - f"「{DUAL_CHECK_FIELD_LABEL['on_reject']}」设为「判为失败」,运行已停止;" + f"「{DUAL_CHECK_FIELD_LABEL['on_reject']}」设为「{DUAL_CHECK_ON_REJECT_LABEL['fail']}」,运行已停止;" "请按审核意见修正取数或口径后重新运行") return _finish(state, ctx, card_id, card, checked, status=RESOLVED, gaps=gaps, choices=choices, note=note, artifact=review["card_artifact"], catalog=catalog) diff --git a/backend/app/engine/schema.py b/backend/app/engine/schema.py index 10c945d3..6a607845 100644 --- a/backend/app/engine/schema.py +++ b/backend/app/engine/schema.py @@ -9,6 +9,13 @@ from typing import Any, Iterable, Literal from pydantic import BaseModel, Field, field_validator, model_validator from app.engine.labels import TYPE_LABEL, field_label, judge_label, option_label, q, type_label # noqa: F401 +# 双盲复算、反方审稿的叫法在 labels.py;校验、执行器和老的调用方照旧从这里取 +from app.engine.labels import ( # noqa: F401 + CHALLENGE_LABEL, + CHALLENGE_ON_REFUTED_LABEL, + DUAL_CHECK_FIELD_LABEL, + DUAL_CHECK_ON_REJECT_LABEL, +) class NodeType(StrEnum): @@ -425,15 +432,9 @@ def _check_report_sources(node: GraphNode, spec: GraphSpec, result: ValidationRe node_id=node.id, field="metrics_from", code=code) -#: 双盲复算节点各配置项在检查器上的叫法(前端 nodeDefs 的 dual_check 字段)。按 labels.py 的规矩应放在那里, -#: 那个文件不归双盲复算的工作包,先放在校验旁边,节点执行时报错也从这里取 -DUAL_CHECK_FIELD_LABEL = { - "card": "口径卡", "metrics": "复算的指标", "definitions": "口径说明", "provider": "盲算模型 · 接入", - "model": "盲算模型", "tolerance": "允许的相对误差", "on_reject": "未采用原流程时", "enabled": "启用双盲复算", -} -#: 审核人没有采用原流程的结果(选了盲算,或者判断两边都不对)时:degrade 计入缺口、出具降档(默认)/ fail 判为失败 +#: 审核人没有采用原流程的结果(选了盲算,或者判断两边都不对)时:degrade 计入缺口、出具降档(默认)/ fail 判为失败。 +#: 各配置项和这两个取值的叫法(DUAL_CHECK_FIELD_LABEL、DUAL_CHECK_ON_REJECT_LABEL)在 labels.py DUAL_CHECK_ON_REJECT = ("degrade", "fail") -DUAL_CHECK_ON_REJECT_LABEL = {"degrade": "计入缺口、出具降档", "fail": "判为失败"} def dual_check_card(spec: GraphSpec, node: GraphNode) -> tuple[str | None, str | None]: @@ -721,17 +722,9 @@ def _check_report_from(node: GraphNode, report_from: Any, spec: GraphSpec, # 反方审稿节点(engine/nodes/challenge.py):审哪份报告、各项配置 # -------------------------------------------------------------------------- -#: 反方审稿节点各项配置在界面上的叫法:报错写叫法,不写键名 -CHALLENGE_LABEL = { - "report": "要审的报告", "max_claims": "最多审几句", "max_queries": "每句最多查询次数", - "timeout_s": "每句时长上限(秒)", "provider": "反方模型 · 接入", "model": "反方模型", - "on_refuted": "结论被推翻时", "enabled": "启用", -} -#: 结论被推翻时怎么办:degrade 计入出具缺口、出具降档(默认)/ flag 只标注 / fail 判节点失败、运行停止 +#: 结论被推翻时怎么办:degrade 计入出具缺口、出具降档(默认)/ flag 只标注 / fail 判节点失败、运行停止。 +#: 各项配置和这三个取值在界面上的叫法(CHALLENGE_LABEL、CHALLENGE_ON_REFUTED_LABEL)在 labels.py:报错写叫法,不写键名 CHALLENGE_ON_REFUTED = ("degrade", "flag", "fail") -#: 选项的叫法和检查器下拉一致(lib/terms.ts 的 CHALLENGE_TEXT.onRefuted),只去掉「(默认)」:报错让人去改某一项时, -#: 人在下拉里找得到这个名字 -CHALLENGE_ON_REFUTED_LABEL = {"degrade": "出具降档", "flag": "只标注", "fail": "报错并停止运行"} def _challenge_field(key: str) -> str: diff --git a/backend/tests/test_verify_labels.py b/backend/tests/test_verify_labels.py new file mode 100644 index 00000000..cdcc5e59 --- /dev/null +++ b/backend/tests/test_verify_labels.py @@ -0,0 +1,45 @@ +"""双盲复算、反方审稿各项配置的叫法放在 engine/labels.py(和别的节点一样),校验、执行报错和通用的 field_label / +option_label 取的是同一份。 + +以前双盲复算的字段叫法写在 schema.py 的 DUAL_CHECK_FIELD_LABEL 里(那时 labels.py 不归双盲复算的工作包),反方审稿的 +写在 schema.py 的 CHALLENGE_LABEL 里:发布门禁、升级预览这些走 field_label 的地方拿不到它们,说成通用的「模型」「启用」, +和检查器上的「盲算模型」「启用双盲复算」对不上。和检查器(lib/terms.ts 的 DUAL_CHECK_TEXT.field、CHALLENGE_TEXT.field +与下拉选项)逐项比对在 check-verify、check-challenge 里做。 +""" +from __future__ import annotations + +from app.engine import labels, schema + + +def test_the_field_labels_live_in_labels_and_schema_reuses_them(): + assert schema.DUAL_CHECK_FIELD_LABEL is labels.DUAL_CHECK_FIELD_LABEL + assert schema.CHALLENGE_LABEL is labels.CHALLENGE_LABEL + assert schema.DUAL_CHECK_ON_REJECT_LABEL is labels.DUAL_CHECK_ON_REJECT_LABEL + assert schema.CHALLENGE_ON_REFUTED_LABEL is labels.CHALLENGE_ON_REFUTED_LABEL + # 每个可选的取值都有叫法,叫法里没有多出来的取值 + assert set(labels.DUAL_CHECK_ON_REJECT_LABEL) == set(schema.DUAL_CHECK_ON_REJECT) + assert set(labels.CHALLENGE_ON_REFUTED_LABEL) == set(schema.CHALLENGE_ON_REFUTED) + + +def test_field_label_names_the_fields_of_the_two_nodes_like_the_inspector(): + assert labels.field_label("model", "dual_check") == "盲算模型" + assert labels.field_label("provider", "dual_check") == "盲算模型 · 接入" + assert labels.field_label("enabled", "dual_check") == "启用双盲复算" + assert labels.field_label("on_reject", "dual_check") == "未采用原流程时" + assert labels.field_label("model", "challenge") == "反方模型" + assert labels.field_label("on_refuted", "challenge") == "结论被推翻时" + assert labels.field_label("max_claims", "challenge") == "最多审几句" + # 别的节点上的同名字段照旧 + assert labels.field_label("model", "agent") == "模型" + for key, words in labels.DUAL_CHECK_FIELD_LABEL.items(): + assert labels.field_label(key, "dual_check") == words, key + for key, words in labels.CHALLENGE_LABEL.items(): + assert labels.field_label(key, "challenge") == words, key + + +def test_option_label_uses_the_short_words_of_the_dropdowns(): + assert labels.option_label("on_refuted", "degrade") == "出具降档" + assert labels.option_label("on_refuted", "flag") == "只标注" + assert labels.option_label("on_refuted", "fail") == "报错并停止运行" + assert labels.option_label("on_reject", "degrade") == "计入缺口、出具降档" + assert labels.option_label("on_reject", "fail") == "判为失败"