Files
YG_FT/frontend/scripts/regression-data-process-wizard.mjs
2026-07-30 16:54:00 +08:00

1480 lines
103 KiB
JavaScript
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
import assert from 'node:assert/strict'
import { existsSync } from 'node:fs'
import { readFile } from 'node:fs/promises'
import { fileURLToPath } from 'node:url'
import path from 'node:path'
import { parse as parseSfc } from '@vue/compiler-sfc'
const scriptDir = path.dirname(fileURLToPath(import.meta.url))
const viewPath = path.resolve(scriptDir, '../src/views/data-process/DataProcessCreateView.vue')
const createDir = path.resolve(scriptDir, '../src/views/data-process/create')
const confirmDialogPath = path.resolve(scriptDir, '../src/components/AppConfirmDialog.vue')
const appHeaderPath = path.resolve(scriptDir, '../src/components/AppHeader.vue')
const layoutPath = path.resolve(scriptDir, '../src/layouts/MainLayout.vue')
const apiPath = path.resolve(scriptDir, '../src/api/modules/dataProcess.ts')
const contractTypesPath = path.resolve(scriptDir, '../src/types/dataProcess.ts')
const requestPath = path.resolve(scriptDir, '../src/api/request.ts')
const authStorePath = path.resolve(scriptDir, '../src/stores/auth.ts')
const modelsStorePath = path.resolve(scriptDir, '../src/stores/models.ts')
const routerPath = path.resolve(scriptDir, '../src/router/index.ts')
const sessionActivityPath = path.resolve(scriptDir, '../src/utils/sessionActivity.ts')
const sourceUploadWorkerPath = path.join(createDir, 'useDataProcessSourceUpload.ts')
const regenerationPath = path.join(createDir, 'useDataProcessRegeneration.ts')
const generationStepPath = path.join(createDir, 'GenerationStep.vue')
const viewSource = await readFile(viewPath, 'utf8')
const layoutSource = await readFile(layoutPath, 'utf8')
const [stateSource, generationSource, previewBuildSource, sourceUploadWorkerSource, regenerationSource, viewStyleSource, apiSource, contractTypesSource] = await Promise.all([
readFile(path.join(createDir, 'dataProcessCreateState.ts'), 'utf8'),
readFile(path.join(createDir, 'useDataProcessGeneration.ts'), 'utf8'),
readFile(path.join(createDir, 'useDataProcessPreviewBuild.ts'), 'utf8'),
readFile(sourceUploadWorkerPath, 'utf8'),
readFile(regenerationPath, 'utf8'),
readFile(path.join(createDir, 'data-process-create.scss'), 'utf8'),
readFile(apiPath, 'utf8'),
readFile(contractTypesPath, 'utf8'),
])
const implementationSource = [viewSource, stateSource, generationSource, previewBuildSource, sourceUploadWorkerSource, regenerationSource].join('\n')
const [requestSource, authStoreSource, modelsStoreSource, routerSource, sessionActivitySource] = await Promise.all([
readFile(requestPath, 'utf8'),
readFile(authStorePath, 'utf8'),
readFile(modelsStorePath, 'utf8'),
readFile(routerPath, 'utf8'),
readFile(sessionActivityPath, 'utf8'),
])
assert.match(requestSource, /if \(res\.code === 0\) \{[\s\S]*?touchSessionActivity\(\)/, '成功 API 请求没有刷新会话活跃时间')
assert.match(authStoreSource, /const loginTime = sessionActivityTime/, '认证状态没有共享请求层的会话活跃时间')
assert.match(authStoreSource, /if \(currentUser\.value\) touchSessionActivity\(\)/, '会话续期仍可能在长任务结束后失效')
assert.match(routerSource, /const auth = useAuthStore\(\)[\s\S]*?auth\.syncSession\(\)[\s\S]*?if \(!auth\.isLoggedIn\)/, '路由守卫没有在登录判断前同步 API 活跃时间')
assert.match(sessionActivitySource, /export function touchSessionActivity\(\)/, '缺少统一会话活跃续期函数')
assert.ok(existsSync(confirmDialogPath), '缺少公共确认弹窗组件 AppConfirmDialog')
const confirmDialogSource = await readFile(confirmDialogPath, 'utf8')
const appHeaderSource = await readFile(appHeaderPath, 'utf8')
for (const marker of ['<Teleport to="body">', 'role="alertdialog"', ':aria-modal="true"', 'handleKeydown', 'Escape']) {
assert.ok(confirmDialogSource.includes(marker), `公共确认弹窗缺少可访问性能力:${marker}`)
}
assert.match(confirmDialogSource, /min-height:\s*44px/, '公共确认弹窗按钮触控区域不足 44px')
assert.match(confirmDialogSource, /focus\(\)/, '公共确认弹窗打开后没有管理键盘焦点')
assert.match(confirmDialogSource, /defineExpose\(\{ open \}\)/, '公共确认弹窗没有暴露 Promise 式 open API')
assert.match(confirmDialogSource, /width:\s*min\(480px,\s*100%\)/, '企业级确认弹窗宽度应保持紧凑的 480px')
assert.match(confirmDialogSource, /border-radius:\s*8px/, '企业级确认弹窗应使用克制的 8px 圆角')
assert.doesNotMatch(confirmDialogSource, /backdrop-filter/, '企业级确认弹窗不应使用装饰性背景模糊')
assert.ok(confirmDialogSource.includes('app-confirm-header'), '企业级确认弹窗缺少独立标题栏')
assert.match(confirmDialogSource, /\.app-confirm-button\s*\{[\s\S]*?height:\s*34px/, '桌面端操作按钮应使用紧凑的 34px 高度')
assert.match(confirmDialogSource, /@media \(max-width: 520px\)[\s\S]*?\.app-confirm-button\s*\{[\s\S]*?min-height:\s*44px/, '移动端操作按钮仍需保留 44px 触控高度')
assert.match(viewSource, /import AppConfirmDialog from '@\/components\/AppConfirmDialog\.vue'/, '创建页没有接入公共确认弹窗')
assert.match(viewSource, /<AppConfirmDialog/, '创建页模板缺少公共确认弹窗实例')
assert.match(viewSource, /onBeforeRouteLeave\(async \(to\) =>/, '路由离开守卫没有根据目标路由异步保存或确认')
assert.doesNotMatch(viewSource, /window\.confirm|ElMessageBox/, '创建页仍在使用系统或 Element Plus 确认框')
assert.match(routerSource, /path:\s*'data-process\/:id\/regenerate'[\s\S]*?name:\s*'data-process-regenerate'[\s\S]*?DataProcessCreateView\.vue/, '重新生成路由没有复用创建向导')
for (const routeName of ['data-process-create', 'data-process-regenerate', 'data-process-workflow']) {
assert.match(
routerSource,
new RegExp(`name:\\s*['"]${routeName}['"][\\s\\S]*?backRouteName:\\s*['"]data-process['"]`),
`数据处理向导路由 ${routeName} 没有声明顶部返回列表的命名路由`,
)
}
assert.match(
appHeaderSource,
/const backRouteName = route\.meta\.backRouteName[\s\S]*?router\.push\(\{ name: backRouteName \}\)[\s\S]*?router\.back\(\)/,
'顶部返回没有优先使用路由 meta 中的命名路由',
)
const persistWorkspaceForStepStart = viewSource.indexOf('async function persistWorkspaceForStep(targetStep: StepId)')
const persistWorkspaceBeforeLeaveStart = viewSource.indexOf('async function persistWorkspaceBeforeLeave()', persistWorkspaceForStepStart)
const persistWorkspaceEnd = viewSource.indexOf('onBeforeRouteLeave(', persistWorkspaceBeforeLeaveStart)
assert.ok(
persistWorkspaceForStepStart >= 0
&& persistWorkspaceBeforeLeaveStart > persistWorkspaceForStepStart
&& persistWorkspaceEnd > persistWorkspaceBeforeLeaveStart,
'已创建任务离开向导前没有持久化当前工作区',
)
const persistWorkspaceForStepSource = viewSource.slice(persistWorkspaceForStepStart, persistWorkspaceBeforeLeaveStart)
for (const persistenceAction of ['saveTaskConfiguration', 'syncPreviewChanges', 'persistResultChanges', 'persistWorkflowStep']) {
assert.ok(persistWorkspaceForStepSource.includes(persistenceAction), `已创建任务离开前缺少保存动作:${persistenceAction}`)
}
const persistWorkspaceBeforeLeaveSource = viewSource.slice(persistWorkspaceBeforeLeaveStart, persistWorkspaceEnd)
assert.match(
persistWorkspaceBeforeLeaveSource,
/persistWorkspaceForStep\(currentStepId\.value\)/,
'无参离开包装没有按当前步骤保存工作区',
)
assert.match(
viewSource,
/async function saveTaskConfiguration\(\)[\s\S]*?updateDataProcessTask\(taskId\.value, taskPayload\(\)\)[\s\S]*?createDataProcessTask\(taskPayload\(\)\)/,
'任务配置保存没有覆盖已创建更新和首次创建',
)
const routeLeaveStart = viewSource.indexOf('onBeforeRouteLeave(')
const routeLeaveEnd = viewSource.indexOf('async function initializeExistingWorkflow()', routeLeaveStart)
assert.ok(routeLeaveStart >= 0 && routeLeaveEnd > routeLeaveStart, '创建向导缺少路由离开守卫')
const routeLeaveSource = viewSource.slice(routeLeaveStart, routeLeaveEnd)
assert.match(
routeLeaveSource,
/to\.name === 'data-process'[\s\S]*?taskId\.value[\s\S]*?await persistWorkspaceBeforeLeave\(\)[\s\S]*?allowLeave = true[\s\S]*?return true/,
'已创建任务返回列表时没有自动保存并直接放行',
)
const persistedLeaveBranch = routeLeaveSource.match(
/if \([^)]*to\.name === 'data-process'[\s\S]*?\) \{([\s\S]*?)return true\s*\}/,
)?.[1] || ''
assert.ok(persistedLeaveBranch, '无法识别已创建任务返回列表的放行分支')
assert.doesNotMatch(persistedLeaveBranch, /confirmDialogRef|confirmDialog|confirm\(/, '已创建任务返回列表不应再弹出离开确认')
assert.match(regenerationSource, /route\.name === 'data-process-regenerate'[\s\S]*?route\.params\.id/, '创建向导没有从路由参数识别重新生成来源任务')
assert.match(viewSource, /const currentStep = ref\(0\)/, '重新生成必须从向导第一步开始')
assert.match(viewSource, /const WIZARD_STEPS = \[/, '向导步骤尚未改为固定常量')
for (const title of ['创建任务', '大模型选择', '上传文件', '数据预览', '开始生成', '结果编辑与保存']) {
assert.ok(viewSource.includes(`title: '${title}'`), `缺少固定步骤:${title}`)
}
assert.match(
viewSource,
/\{ id: 'create',[\s\S]*?\{ id: 'model',[\s\S]*?\{ id: 'upload',[\s\S]*?\{ id: 'preview',[\s\S]*?\{ id: 'generate',[\s\S]*?\{ id: 'results'/,
'六步向导顺序必须为创建任务、大模型选择、上传文件、数据预览、开始生成、结果编辑与保存',
)
assert.doesNotMatch(viewSource, /steps\s*=\s*computed|all\.filter/, '步骤仍根据处理类型动态增减')
assert.match(
viewStyleSource,
/@media \(max-width: 1100px\)[\s\S]*?\.step-title\s*\{[\s\S]*?display:\s*none[\s\S]*?\.step-item\.is-active \.step-title\s*\{[\s\S]*?display:\s*block/,
'六步向导在中等宽度下没有收起非当前步骤标题',
)
assert.equal(existsSync(path.join(createDir, 'useDataProcessDraft.ts')), false, '不应保留新建任务草稿模块')
assert.doesNotMatch(
viewSource,
/useDataProcessDraft|persistDraft|restoreDraft|restoringDraft|localStorage\.(?:setItem|getItem)/,
'新建任务不应保存或恢复草稿',
)
assert.match(viewSource, /localStorage\.removeItem\('yg-data-process-create-draft'\)/, '进入新建页时应清理遗留草稿')
assert.ok(viewSource.split('\n').length < 1200, 'DataProcessCreateView 拆分后仍超过 1200 行')
assert.match(viewSource, /useDataProcessGeneration\(\{/, '生成流程没有拆分到独立 composable')
const expectedComponents = [
'TaskSetupStep.vue',
'ModelSelectionStep.vue',
'SourceUploadStep.vue',
'PreviewCompareStep.vue',
'GenerationStep.vue',
'ResultEditorStep.vue',
]
for (const component of expectedComponents) {
assert.ok(existsSync(path.join(createDir, component)), `缺少步骤组件:${component}`)
assert.ok(viewSource.includes(component.replace('.vue', '')), `父页面未使用:${component}`)
}
const typesPath = path.join(createDir, 'types.ts')
const modelPath = path.join(createDir, 'previewModel.ts')
const pdfViewerPath = path.join(createDir, 'PdfSourceViewer.vue')
const officeViewerPath = path.join(createDir, 'OfficeSourceViewer.vue')
const resultEditorPath = path.join(createDir, 'ResultEditorStep.vue')
assert.ok(existsSync(typesPath), '缺少向导类型定义')
assert.ok(existsSync(modelPath), '缺少来源映射模型')
assert.ok(existsSync(pdfViewerPath), '缺少 PDF 原文件预览组件')
assert.ok(existsSync(officeViewerPath), '缺少 Word/XLSX 原文件预览组件')
const [typesSource, modelSource, previewSource, pdfViewerSource, officeViewerSource, resultEditorSource, generationStepSource] = await Promise.all([
readFile(typesPath, 'utf8'),
readFile(modelPath, 'utf8'),
readFile(path.join(createDir, 'PreviewCompareStep.vue'), 'utf8'),
readFile(pdfViewerPath, 'utf8'),
readFile(officeViewerPath, 'utf8'),
readFile(resultEditorPath, 'utf8'),
readFile(generationStepPath, 'utf8'),
])
for (const field of ['sourceStart', 'sourceEnd', 'originalContent', 'editedContent']) {
assert.ok(typesSource.includes(field), `PreviewItem 缺少字段:${field}`)
}
assert.match(typesSource, /sourceFileId/, 'PreviewItem 缺少来源文件标识')
assert.match(typesSource, /sourceLocator\?: PreviewSourceLocator/, 'PreviewItem 缺少结构化来源定位契约')
assert.match(typesSource, /headingPath\?: string\[\]/, 'PreviewItem 缺少非结构化标题路径')
assert.match(typesSource, /PreviewSourceLocatorKind = 'json' \| 'jsonl' \| 'csv' \| 'xlsx'/, '前端来源定位 kind 未使用明确联合类型')
assert.match(contractTypesSource, /DataProcessSourceLocatorKind = 'json' \| 'jsonl' \| 'csv' \| 'xlsx'/, 'API 来源定位 kind 未使用明确联合类型')
for (const field of ['kind', 'record_index', 'start_line', 'end_line', 'source_start', 'source_end', 'json_pointer', 'sheet_index', 'sheet_name', 'row_number', 'sheet_record_index']) {
assert.ok(typesSource.includes(field), `PreviewSourceLocator 缺少字段:${field}`)
assert.ok(contractTypesSource.includes(field), `后端来源定位契约缺少字段:${field}`)
}
assert.match(contractTypesSource, /source_locator\?: DataProcessSourceLocator/, '质量信息缺少来源定位契约')
assert.match(contractTypesSource, /heading_path\?: string\[\]/, '质量信息缺少标题路径契约')
assert.match(viewSource, /const sourceLocator = item\.quality_score\?\.source_locator/, '预览映射丢失来源定位')
assert.match(viewSource, /sourceStart:\s*item\.source_start\s*\?\?\s*sourceLocator\?\.source_start/, 'JSON locator 的字符起点没有映射到预览项')
assert.match(viewSource, /sourceEnd:\s*item\.source_end\s*\?\?\s*sourceLocator\?\.source_end/, 'JSON locator 的字符终点没有映射到预览项')
assert.match(viewSource, /sourceStartLine:\s*item\.source_start_line\s*\?\?\s*sourceLocator\?\.start_line/, 'JSON locator 的起始行没有映射到预览项')
assert.match(viewSource, /sourceEndLine:\s*item\.source_end_line\s*\?\?\s*sourceLocator\?\.end_line/, 'JSON locator 的结束行没有映射到预览项')
assert.match(viewSource, /headingPath:[\s\S]*?item\.quality_score\?\.heading_path/, '预览映射丢失标题路径')
assert.match(typesSource, /export type StepId = 'create' \| 'model' \| 'upload' \| 'preview' \| 'generate' \| 'results'/, '步骤类型缺少独立大模型选择步骤')
assert.match(modelSource, /export function sourceLineWindow/, '缺少有界源文件行窗口函数')
assert.match(modelSource, /maxLines:\s*number/, '源文件行窗口缺少最大渲染行数参数')
assert.doesNotMatch(modelSource, /\.split\(\s*['"]\\n['"]\s*\)/, '源文件行窗口仍会先对全文 split')
assert.match(modelSource, /lines\.length < limit/, '源文件行扫描没有受最大行数约束')
assert.match(modelSource, /unicodeCodePointLength/, '源文件字符偏移未与后端 Unicode code point 计数保持一致')
assert.match(modelSource, /export function sourceLineNumberAtOffset/, '字符偏移缺少无数组的行号解析函数')
const manualPreviewHelperStart = modelSource.indexOf('export function isManualPreviewItem(')
const manualPreviewHelperEnd = modelSource.indexOf('\n}', manualPreviewHelperStart)
assert.ok(manualPreviewHelperStart >= 0, '缺少统一的手动预览项判定函数')
const manualPreviewHelperSource = modelSource.slice(manualPreviewHelperStart, manualPreviewHelperEnd + 2)
for (const field of ['status', 'originalContent', 'sourceStart', 'sourceEnd', 'sourceStartLine', 'sourceEndLine', 'sourcePages', 'sourceLocator']) {
assert.ok(manualPreviewHelperSource.includes(field), `手动预览项判定缺少来源字段:${field}`)
}
assert.doesNotMatch(modelSource, /buildPreviewItems/, '前端不应保留与后端重复的本地切片算法')
assert.match(viewSource, /selectedPreviewFileId/, '父页面缺少当前预览文件状态')
const previewBuildBindingStart = viewSource.indexOf('useDataProcessPreviewBuild()')
assert.ok(previewBuildBindingStart >= 0, '父页面没有接入后台切分 composable')
const previewBuildBindingSource = viewSource.slice(Math.max(0, previewBuildBindingStart - 160), previewBuildBindingStart + 40)
for (const action of ['startPreviewBuild', 'resumePreviewBuild', 'stopPreviewPolling']) {
assert.ok(previewBuildBindingSource.includes(action), `父页面缺少后台切分能力:${action}`)
}
assert.match(viewSource, /backend-pipeline-v4/, '切分管线版本未升级,旧预览缓存可能被误用')
assert.match(viewSource, /getDataProcessPreview\(taskId\.value,\s*\{ page:\s*1, page_size:\s*500 \}\)/, '预览构建后没有分页读取后端数据')
assert.doesNotMatch(viewSource, /buildPreviewItems\(/, '创建向导仍在本地构建集成预览数据')
for (const marker of [
'preview-workspace',
'source-viewer',
'source-line',
'is-highlighted',
'preview-item',
'preview-editor',
'scrollIntoView',
]) {
assert.ok(previewSource.includes(marker), `第四步缺少结构或行为:${marker}`)
}
assert.match(previewSource, /sourceStart/, '第四步未使用来源起始偏移')
assert.match(previewSource, /sourceEnd/, '第四步未使用来源结束偏移')
const lineRangeStart = previewSource.indexOf('function lineRange(item: PreviewItem)')
const lineRangeEnd = previewSource.indexOf('\n}', lineRangeStart)
const lineRangeSource = previewSource.slice(lineRangeStart, lineRangeEnd + 2)
assert.match(lineRangeSource, /isManualPreviewItem\(item\)[\s\S]*?手动新增,无源文件定位/, '来源标签仍会把缺少行偏移的正常记录误判为手动新增')
assert.match(lineRangeSource, /props\.processType === 'unstructured'[\s\S]*?来源:源文件记录/, '结构化来源记录缺少无行偏移时的准确标签')
assert.doesNotMatch(lineRangeSource, /sourceStartLine == null[^\n]*手动新增/, '来源标签仍直接以缺少行号判定手动新增')
assert.match(lineRangeSource, /sheet_name[\s\S]*?row_number[\s\S]*?来源:\$\{sheet\} · 第 \$\{locator\.row_number\} 行/, 'XLSX 来源标签没有展示工作表和物理行号')
assert.match(lineRangeSource, /json_pointer[\s\S]*?JSON 路径/, 'JSON 来源标签没有展示 JSON 路径')
assert.match(lineRangeSource, /locator\?\.kind === 'json'[\s\S]*?JSON 根对象/, 'JSON 根对象来源标签被空 JSON Pointer 错误降级')
assert.match(lineRangeSource, /locatedLines[\s\S]*?第 \$\{locatedLines\.start\}[\s\S]*?locatedLines\.end/, 'JSONL/CSV 来源标签没有展示行范围')
assert.match(lineRangeSource, /headingPath[\s\S]*?章节:/, '非结构化来源标签没有合并标题路径')
assert.match(previewSource, /sourceLocator\?\.start_line[\s\S]*?sourceLocator\?\.end_line/, '文本预览没有优先使用后端行号定位')
assert.match(previewSource, /sourceLocator\?\.source_start\s*\?\?\s*item\.sourceStart/, '文本预览没有优先使用 locator 字符起点')
assert.match(previewSource, /sourceLocator\?\.source_end\s*\?\?\s*item\.sourceEnd/, '文本预览没有优先使用 locator 字符终点')
assert.match(previewSource, /data-line-number="line\.number"/, '文本预览行缺少稳定行号定位标识')
assert.match(previewSource, /isLineHighlighted\(line\.number, line\.start, line\.end\)/, '文本预览没有按物理行号高亮')
assert.match(previewSource, /querySelector<HTMLElement>\(`\[data-line-number=/, '选中记录后没有按物理行号滚动定位')
assert.match(previewSource, /const SOURCE_LINE_RENDER_LIMIT = 240/, '源文件查看器缺少安全渲染上限')
assert.match(previewSource, /const SOURCE_LINE_CHARACTER_LIMIT = 4_000/, '源文件查看器缺少单行字符渲染上限')
assert.match(previewSource, /sourceLineWindow\([\s\S]*?SOURCE_LINE_RENDER_LIMIT/, '源文件查看器没有使用有界行窗口')
assert.match(previewSource, /SOURCE_LINE_RENDER_LIMIT,[\s\S]*?SOURCE_LINE_CHARACTER_LIMIT,[\s\S]*?selectedSourceLine\.value,[\s\S]*?selectedSourceOffset\.value/, '单行超大 JSON 没有围绕选中来源构建字符窗口')
assert.match(previewSource, /sourceWindowStartLine/, '源文件查看器缺少窗口起始行状态')
assert.match(previewSource, /showPreviousSourceWindow[\s\S]*?showNextSourceWindow/, '源文件查看器缺少前后窗口导航')
assert.match(previewSource, /sourceLineNumberAtOffset\(props\.sourceText/, '仅有字符偏移时没有解析目标物理行')
assert.match(previewSource, /filterable/, '文件选择器必须可搜索')
assert.match(previewSource, /当前文件/, '预览缺少当前文件切换器')
assert.doesNotMatch(previewSource, /located-badge|sync-label|已定位到/, '源文件栏不应显示冗余定位提示')
assert.match(previewSource, /const PREVIEW_PAGE_SIZE = 10/, '切片列表每页应展示 10 条')
assert.match(previewSource, /const pagedItems = computed/, '切片列表缺少分页数据')
assert.match(previewSource, /v-for="item in pagedItems"/, '切片列表没有使用分页数据')
assert.match(previewSource, /<el-pagination[\s\S]*:page-size="PREVIEW_PAGE_SIZE"/, '切片列表缺少分页控件')
assert.match(previewSource, /height:\s*clamp\(560px,\s*calc\(100vh - 370px\),\s*720px\)/, '预览工作区高度不足以展示切片正文')
assert.match(previewSource, /const editingItemId = ref<string \| null>\(null\)/, '缺少切片编辑模式状态')
assert.match(previewSource, /const editorDraft = ref\(''\)/, '缺少编辑临时草稿')
assert.match(previewSource, /function openEditor\(item: PreviewItem\)/, '列表缺少打开切片编辑器的动作')
assert.match(previewSource, /function closeEditor\(\)/, '编辑器缺少返回列表的动作')
assert.match(previewSource, /function saveEditor\(\)/, '编辑器缺少保存动作')
assert.match(previewSource, /<template v-if="!editingItem">[\s\S]*?<template v-else>/, '切片列表与编辑器必须互斥展示')
assert.match(previewSource, /fa-pencil/, '切片列表缺少铅笔编辑按钮')
assert.match(previewSource, /fa-trash-o/, '切片列表缺少垃圾桶删除按钮')
assert.match(previewSource, /class="preview-item"[\s\S]*?@click="selectItem\(item\.id\)"/, '点击切片行必须更新当前选中切片')
assert.match(previewSource, /v-model="editorDraft"/, '编辑器必须绑定临时草稿')
assert.match(previewSource, />取消<\/el-button>/, '编辑器缺少取消按钮')
assert.doesNotMatch(previewSource, /返回列表/, '编辑器不应同时显示返回列表和取消两个相同作用的按钮')
assert.match(previewSource, />保存修改<\/el-button>/, '编辑器缺少保存修改按钮')
assert.match(previewSource, /\.editor-actions\s*\{[\s\S]*?justify-content:\s*flex-end/, '取消和保存按钮必须在编辑器右侧对齐')
assert.doesNotMatch(previewSource, /item-token|item-status|modifiedOnly|仅看已修改/, '切片列表不应再显示 Token 或修改状态')
assert.match(previewSource, /\.preview-editor\s*\{[\s\S]*?flex:\s*1 1 auto[\s\S]*?overflow-y:\s*auto/, '编辑模式必须占据右侧剩余区域并可滚动')
assert.match(previewSource, /@media \(max-width: 900px\)/, '第四步缺少窄屏上下布局')
assert.match(previewSource, /<PdfSourceViewer[\s\S]*v-if="isPdfSource"/, 'PDF 文件没有切换到原文件查看组件')
assert.match(previewSource, /:selected-item="selectedItem \?\? null"/, 'PDF 查看组件没有接收当前选中切片')
assert.match(previewSource, /<OfficeSourceViewer[\s\S]*v-else-if="isOfficeSource"/, 'Word/XLSX 没有切换到专用原文件查看组件')
assert.match(previewSource, /const isOfficeSource = computed[\s\S]*?'docx', 'xlsx'/, 'Word/XLSX 文件类型分流不完整')
assert.match(previewSource, /:data-preview-id="item\.id"/, '切片行缺少稳定的交互定位标识')
assert.match(previewSource, /<div v-else ref="sourceViewerRef" class="source-viewer"/, '非 PDF 文件没有保留文本预览')
assert.match(pdfViewerSource, /getDataProcessSourceRawUrl/, 'PDF 查看组件没有使用受控原文件地址')
assert.match(pdfViewerSource, /getDataProcessPdfPages/, 'PDF 查看组件没有读取页码与全文偏移映射')
assert.match(pdfViewerSource, /pdfjs-dist/, 'PDF 查看组件没有使用可控的 PDF.js 渲染器')
assert.match(pdfViewerSource, /TextLayer/, 'PDF 查看组件没有渲染可定位的 PDF 文字层')
assert.match(pdfViewerSource, /page\.streamTextContent\(\)/, 'PDF 文字层没有使用兼容 WebKit 的流式读取方式')
assert.doesNotMatch(pdfViewerSource, /page\.getTextContent\(\)/, 'PDF 文字层仍依赖 WebKit 不完整支持的 ReadableStream 异步迭代')
assert.match(pdfViewerSource, /is-slice-highlighted/, 'PDF 查看组件没有实现选中切片高亮')
assert.match(pdfViewerSource, /:data-page-number="currentPage"/, 'PDF 查看组件缺少当前物理页标识')
assert.doesNotMatch(pdfViewerSource, /<iframe/, 'PDF 查看组件不应继续使用无法控制高亮的浏览器 iframe')
assert.match(pdfViewerSource, /:aria-label="`PDF 预览:\$\{fileName\}`"/, 'PDF 查看器缺少可访问标题')
assert.match(apiSource, /getDataProcessSourceRawUrl/, '前端 API 缺少 PDF 原文件预览地址')
assert.match(apiSource, /getDataProcessPdfPages/, '前端 API 缺少 PDF 页码映射接口')
assert.match(apiSource, /source-files\/\$\{encodeURIComponent\(fileId\)\}\/pdf-pages/, 'PDF 页码映射接口地址不正确')
assert.match(apiSource, /getDataProcessOfficePreview/, '前端 API 缺少 Word/XLSX 预览接口')
assert.match(apiSource, /source-files\/\$\{encodeURIComponent\(fileId\)\}\/office-preview/, 'Word/XLSX 预览接口地址不正确')
for (const marker of [
'docx-page',
'docx-table',
'xlsx-grid',
'sheet-selector',
'xlsx-pagination',
'getDataProcessOfficePreview',
'is-highlighted',
'打开原文件',
'重试',
]) {
assert.ok(officeViewerSource.includes(marker), `Word/XLSX 预览缺少结构或行为:${marker}`)
}
assert.match(officeViewerSource, /const selectedXlsxLocator = computed/, 'XLSX 查看器没有读取精确来源定位')
assert.match(officeViewerSource, /row\.row_number === locator\.row_number/, 'XLSX 查看器没有按物理行号精确高亮')
assert.match(officeViewerSource, /row\.record_index === locator\.sheet_record_index/, 'XLSX 查看器没有按工作表记录序号精确高亮')
assert.match(officeViewerSource, /Math\.floor\(locator\.sheet_record_index \/ XLSX_PAGE_SIZE\) \* XLSX_PAGE_SIZE/, 'XLSX 查看器没有按记录序号自动计算分页')
assert.match(officeViewerSource, /activeSheetIndex\.value = targetSheet[\s\S]*?pageOffset\.value = targetOffset[\s\S]*?loadPreview\(\)/, '切换记录时 XLSX 查看器没有自动切工作表和分页')
const xlsxHighlightStart = officeViewerSource.indexOf('function xlsxRowHighlighted(')
const xlsxHighlightEnd = officeViewerSource.indexOf('\n}', xlsxHighlightStart)
const xlsxHighlightSource = officeViewerSource.slice(xlsxHighlightStart, xlsxHighlightEnd + 2)
assert.ok(
xlsxHighlightSource.indexOf('locator.row_number') < xlsxHighlightSource.indexOf('selectedRecordKey.value'),
'XLSX 查看器没有把精确定位放在原内容比对 fallback 之前',
)
const taskSetupPath = path.join(createDir, 'TaskSetupStep.vue')
const structuredOptionsPath = path.join(createDir, 'StructuredOptionsPanel.vue')
const unstructuredOptionsPath = path.join(createDir, 'UnstructuredOptionsPanel.vue')
const datasetSplitEditorPath = path.join(createDir, 'DatasetSplitEditor.vue')
const generationOptionsPath = path.join(createDir, 'GenerationOptionsPanel.vue')
const modelSelectionPath = path.join(createDir, 'ModelSelectionStep.vue')
const sourceUploadPath = path.join(createDir, 'SourceUploadStep.vue')
const [
taskSetupSource,
structuredOptionsSource,
unstructuredOptionsSource,
datasetSplitEditorSource,
generationControlSource,
modelSelectionSource,
sourceUploadSource,
] = await Promise.all([
readFile(taskSetupPath, 'utf8'),
readFile(structuredOptionsPath, 'utf8'),
readFile(unstructuredOptionsPath, 'utf8'),
readFile(datasetSplitEditorPath, 'utf8'),
readFile(generationOptionsPath, 'utf8'),
readFile(modelSelectionPath, 'utf8'),
readFile(sourceUploadPath, 'utf8'),
])
const taskSetupFeatureSource = [
taskSetupSource,
structuredOptionsSource,
unstructuredOptionsSource,
datasetSplitEditorSource,
].join('\n')
for (const componentPath of [structuredOptionsPath, unstructuredOptionsPath, datasetSplitEditorPath]) {
assert.ok(existsSync(componentPath), `缺少任务配置拆分组件:${path.basename(componentPath)}`)
}
assert.ok(taskSetupSource.split('\n').length < 800, 'TaskSetupStep 拆分后仍超过 800 行')
assert.match(taskSetupSource, /<StructuredOptionsPanel/, '任务配置没有挂载结构化选项面板')
assert.match(taskSetupSource, /<UnstructuredOptionsPanel/, '任务配置没有挂载非结构化选项面板')
assert.match(structuredOptionsSource, /<DatasetSplitEditor/, '结构化选项没有复用数据集划分编辑器')
assert.match(unstructuredOptionsSource, /<DatasetSplitEditor/, '非结构化选项没有复用数据集划分编辑器')
for (const marker of ['<el-upload', '源数据上传', '数据源配置', 'uploadedFiles']) {
assert.ok(!taskSetupFeatureSource.includes(marker), `第一步仍包含上传职责:${marker}`)
}
assert.match(viewSource, /<ModelSelectionStep\s+[\s\S]*?v-else-if="currentStepId === 'model'"/, '第二步没有挂载独立大模型选择组件')
assert.match(viewSource, /<SourceUploadStep\s+[\s\S]*?v-else-if="currentStepId === 'upload'"/, '第三步没有挂载独立上传组件')
assert.match(viewSource, /if \(currentStepId\.value === 'create'\) return '继续:选择大模型'/, '第一步主按钮没有指向大模型选择')
assert.match(viewSource, /if \(currentStepId\.value === 'model'\) return '继续:上传文件'/, '第二步主按钮没有指向上传文件')
assert.match(
viewSource,
/if \(currentStepId\.value === 'upload'\) \{[\s\S]*?sourceUploading\.value[\s\S]*?'正在上传'[\s\S]*?previewBuilding\.value \? '正在切分' : '继续:数据预览'/,
'第三步主按钮没有依次反映上传、切分状态并指向数据预览',
)
const nextFromCreateStart = viewSource.indexOf('async function nextFromCreate()')
const nextFromModelStart = viewSource.indexOf('async function nextFromModel()', nextFromCreateStart)
const nextFromUploadStart = viewSource.indexOf('function nextFromUpload()', nextFromModelStart)
const selectPreviewFileStart = viewSource.indexOf('function selectPreviewFile(', nextFromUploadStart)
assert.ok(
nextFromCreateStart >= 0 && nextFromModelStart > nextFromCreateStart && nextFromUploadStart > nextFromModelStart,
'缺少创建、大模型选择与上传步骤的独立跳转函数',
)
const nextFromCreateSource = viewSource.slice(nextFromCreateStart, nextFromModelStart)
const nextFromModelSource = viewSource.slice(nextFromModelStart, nextFromUploadStart)
const nextFromUploadSource = viewSource.slice(nextFromUploadStart, selectPreviewFileStart)
assert.match(nextFromCreateSource, /taskSetupRef\.value\?\.validate\(\)/, '创建步骤继续前没有校验任务配置')
assert.match(nextFromCreateSource, /saveTaskConfiguration\(\)[\s\S]*?persistWorkflowStep\('model'\)/, '进入第二步前没有创建任务并持久化步骤')
assert.match(nextFromCreateSource, /goToStep\('model'\)/, '创建步骤校验通过后没有进入大模型选择')
assert.doesNotMatch(nextFromCreateSource, /uploadedFiles|buildDataProcessPreview/, '创建步骤仍在校验文件或提前生成预览')
assert.match(nextFromModelSource, /modelSelectionRef\.value\?\.validate\(\)/, '大模型选择步骤继续前没有校验模型配置')
assert.match(nextFromModelSource, /saveTaskConfiguration\(\)[\s\S]*?persistWorkflowStep\('upload'\)/, '大模型选择完成后没有保存配置并持久化步骤')
assert.match(nextFromModelSource, /goToStep\('upload'\)/, '大模型选择完成后没有进入上传文件')
assert.match(nextFromModelSource, /await modelsStore\.load\(true\)/, '提交任务前没有等待最新模型列表')
assert.ok(nextFromModelSource.indexOf('await modelsStore.load(true)') < nextFromModelSource.indexOf('prepareRegeneration(taskPayload())'), '重新生成可能在模型列表校验前修改服务端')
assert.ok(nextFromModelSource.indexOf('await modelsStore.load(true)') < nextFromModelSource.indexOf('await saveTaskConfiguration()'), '普通任务可能在模型列表校验前修改服务端')
assert.ok(nextFromModelSource.includes('模型列表加载失败,未提交任何修改'), '模型列表加载失败没有明确阻断提交')
assert.ok(nextFromModelSource.includes('原任务使用的模型已删除'), '原任务模型已删除没有要求重新选择')
assert.match(modelsStoreSource, /catch \{[\s\S]*?list\.value = \[\][\s\S]*?loaded\.value = false/, '模型列表请求失败后 loaded 状态仍可能错误保留')
assert.match(nextFromUploadSource, /uploadedFiles\.value\.length === 0/, '上传步骤继续前没有校验源数据')
assert.match(
nextFromUploadSource,
/const pendingFileIds = uploadedFiles\.value[\s\S]*?\.filter\(\(file\) => file\.previewStatus !== 'success' \|\| file\.previewConfigSignature !== configSignature\)[\s\S]*?\.map\(\(file\) => file\.sourceFileId\)[\s\S]*?\.filter\(\(fileId\): fileId is string => Boolean\(fileId\)\)/,
'上传步骤没有仅选择待处理或失败文件,无法跳过成功文件并重试失败文件',
)
assert.match(
nextFromUploadSource,
/await monitorPreviewBuild\(pendingFileIds\)/,
'上传步骤没有通过独立 composable 启动后台切分',
)
assert.match(
previewBuildSource,
/startDataProcessPreview\(taskId,[\s\S]*?replace_existing:\s*true[\s\S]*?source_file_ids:\s*sourceFileIds/,
'切分 composable 没有调用后台启动 API 并传递待处理文件',
)
assert.match(
previewBuildSource,
/function isActive\([\s\S]*?preview_status === 'queued'[\s\S]*?preview_status === 'running'/,
'切分 composable 没有将 queued\/running 识别为活动状态',
)
assert.match(
previewBuildSource,
/while \(isActive\(progress\)[\s\S]*?getDataProcessPreviewProgress\(taskId\)/,
'切分 composable 没有持续轮询 queued\/running 的服务端状态',
)
assert.match(
previewBuildSource,
/async function resumePreviewBuild\([\s\S]*?getDataProcessPreviewProgress\(taskId\)[\s\S]*?pollUntilSettled/,
'重新进入上传步骤时没有接管已在后台运行的切分',
)
if (previewBuildSource.includes('function buildPreviewsByFile')) {
assert.match(
previewBuildSource,
/function buildPreviewsByFile[\s\S]*?startPreviewBuild\(taskId, sourceFileIds/,
'逐文件展示兼容层底层仍必须只启动一个后台切分任务',
)
}
assert.match(
previewBuildSource,
/onBeforeUnmount\(stopPreviewPolling\)/,
'离开页面时应只停止前端轮询,不得终止后台切分',
)
assert.match(
previewBuildSource,
/const pollRun = activePollRun[\s\S]*?await startDataProcessPreview[\s\S]*?pollUntilSettled\(taskId, progress, pollRun/,
'后台切分启动请求返回后必须复用原轮询令牌,避免页面卸载后重新启动轮询',
)
assert.match(
previewBuildSource,
/if \(pollRun !== activePollRun\) return progress/,
'已离开页面的后台切分请求不得重新接管页面轮询',
)
assert.doesNotMatch(previewBuildSource, /buildDataProcessPreview\(/, '新向导不应再依赖同步 preview\/build 请求')
assert.match(
viewSource,
/async function monitorPreviewBuild\([\s\S]*?startPreviewBuild\(taskId\.value, sourceFileIds[\s\S]*?finalProgress\.preview_status === 'completed'[\s\S]*?completePreviewWorkspace\(\)/,
'上传步骤没有启动后台切分或仅在完成后加载工作区',
)
assert.match(
viewSource,
/async function completePreviewWorkspace\(\)[\s\S]*?loadAllPreviewItems\(\)[\s\S]*?failedCount[\s\S]*?return false[\s\S]*?persistWorkflowStep\('preview'\)[\s\S]*?goToStep\('preview'\)/,
'后台切分未完成时向导不得越过第三步',
)
assert.match(viewSource, /function goToStep\(stepId: StepId\)[\s\S]*?WIZARD_STEPS\.findIndex/, '向导跳转没有使用稳定步骤标识')
assert.match(contractTypesSource, /export type DataProcessWorkflowStep = 'create' \| 'model' \| 'upload' \| 'preview' \| 'generate' \| 'results'/, '服务端向导步骤契约不完整')
assert.match(contractTypesSource, /workflow_step\?: DataProcessWorkflowStep/, '任务快照缺少可恢复的 workflow_step')
assert.match(
apiSource,
/export const updateDataProcessWorkflowStep[\s\S]*?\/workflow-step`[\s\S]*?\{ workflow_step: workflowStep \}/,
'前端 API 没有使用独立资源持久化向导步骤',
)
assert.match(
viewSource,
/async function persistWorkflowStep\([\s\S]*?updateDataProcessWorkflowStep\(taskId\.value, step as DataProcessWorkflowStep\)/,
'向导步骤切换没有持久化 workflow_step',
)
for (const workflowStep of ['model', 'upload', 'preview', 'generate', 'results']) {
assert.match(
viewSource,
new RegExp(`persistWorkflowStep\\(['"]${workflowStep}['"]\\)`),
`第 2-6 步推进缺少服务端持久化:${workflowStep}`,
)
}
assert.match(routerSource, /path:\s*'data-process\/:id\/workflow'[\s\S]*?name:\s*'data-process-workflow'[\s\S]*?DataProcessCreateView\.vue/, '运行中任务缺少向导恢复路由')
assert.match(regenerationSource, /route\.name === 'data-process-workflow'/, '创建向导没有识别任务恢复模式')
assert.match(generationSource, /async function resumeGeneration\(\)[\s\S]*?getDataProcessProgress\(taskId\)[\s\S]*?pollGeneration/, '恢复第五步时没有接管后台任务进度')
const workflowInitializationStart = viewSource.indexOf('async function initializeExistingWorkflow()')
const workflowInitializationEnd = viewSource.indexOf('onBeforeUnmount(', workflowInitializationStart)
assert.ok(workflowInitializationStart >= 0 && workflowInitializationEnd > workflowInitializationStart, '缺少现有任务向导初始化流程')
const workflowInitializationSource = viewSource.slice(workflowInitializationStart, workflowInitializationEnd)
assert.match(
workflowInitializationSource,
/sourceTask\.workflow_step[\s\S]*?goToStep/,
'恢复向导时没有优先使用服务端 workflow_step',
)
assert.doesNotMatch(
workflowInitializationSource,
/shouldReturnToUpload|preview_count[\s\S]*?\?\s*'upload'\s*:\s*'generate'/,
'向导恢复不得继续依赖 preview_count 猜测第三或第五步',
)
assert.match(
workflowInitializationSource,
/preview_status[\s\S]*?(?:'queued'[\s\S]*?'running'|'running'[\s\S]*?'queued')[\s\S]*?goToStep\('upload'\)[\s\S]*?monitorPreviewBuild\(sourceFileIds, true\)/,
'切分运行中时没有强制回到第三步并接管后台进度',
)
assert.match(
workflowInitializationSource,
/sourceTask\.status === 'running'[\s\S]*?resumeStep = 'generate'[\s\S]*?resumeGeneration\(\)[\s\S]*?goToStep\(resumeStep\)/,
'生成运行中时没有强制回到第五步并接管后台进度',
)
const startGenerationHandler = viewSource.slice(
viewSource.indexOf('async function handleStartGeneration()'),
viewSource.indexOf('async function persistWorkspaceForStep(', viewSource.indexOf('async function handleStartGeneration()')),
)
assert.match(startGenerationHandler, /await persistWorkflowStep\('generate'\)[\s\S]*?await startGeneration\(\)[\s\S]*?dirty\.value = false/, '开始生成没有持久化第五步或启动真实后台任务')
assert.doesNotMatch(startGenerationHandler, /router\.(?:push|replace)|allowLeave\s*=\s*true/, '开始生成后应停留在第五步,不得自动跳回列表')
assert.match(
generationSource,
/const canReturnFromGeneration = computed\(\(\) => \([\s\S]*?generation\.status === 'idle'[\s\S]*?!generationStarting\.value[\s\S]*?!generationRestoring\.value/,
'第五步返回权限没有区分未启动、启动中和恢复中状态',
)
assert.match(
viewSource,
/:disabled="\(currentStepId === 'generate' && !canReturnFromGeneration\) \|\| previewBuilding \|\| sourceUploading"/,
'第五步尚未启动生成时返回按钮仍被禁用',
)
const handleBackStart = viewSource.indexOf('async function handleBack()')
const handleBackEnd = viewSource.indexOf('\n}', handleBackStart)
const handleBackSource = viewSource.slice(handleBackStart, handleBackEnd + 2)
assert.match(
handleBackSource,
/currentStepId\.value === 'generate' && !canReturnFromGeneration\.value/,
'第五步处理函数仍无条件拦截返回',
)
assert.match(
generationSource,
/async function resumeGeneration\(\)[\s\S]*?generationRestoring\.value = true[\s\S]*?await getDataProcessProgress\(taskId\)[\s\S]*?generationRestoring\.value = false/,
'恢复已启动任务时存在短暂可返回的 idle 窗口',
)
assert.match(viewSource, /const resume = resumeGeneration\(\)[\s\S]*?goToStep\(resumeStep\)[\s\S]*?await resume/, '第五步展示时未先启动恢复锁')
assert.match(
generationSource,
/const generationStarting = ref\(false\)[\s\S]*?generationStarting\.value = true[\s\S]*?generationStarting\.value = false/,
'点击开始生成后到请求启动前没有锁定返回状态',
)
assert.match(
viewSource,
/generation\.status === 'success'[\s\S]*?persistWorkflowStep\('results'\)/,
'只有生成完成后才能持久化进入第六步',
)
assert.match(viewSource, /confirmDataProcessResults\(taskId\.value\)[\s\S]*?name: 'data-process-detail'/, '第六步没有在确认结果后进入正式详情')
assert.doesNotMatch(generationStepSource, /停止生成|emit\(['"]stop['"]\)|stop:\s*\[\]/, '第五步不应保留停止生成入口')
assert.doesNotMatch(viewSource, /@stop=|\bstopGeneration\b/, '创建向导不应绑定删除以外的生成终止动作')
assert.doesNotMatch(generationSource, /\bstopDataProcess\b|async function stopGeneration\b|\bstopGeneration,/, '生成 composable 不应暴露任务终止能力')
assert.match(viewSource, /<ResultEditorStep\s+[\s\S]*?v-else-if="currentStepId === 'results'"/, '结果编辑器必须只在结果步骤渲染')
assert.match(viewSource, /<ResultEditorStep[\s\S]*?:preview-items="previewItems"/, '结果编辑器没有接收已有预览原文')
assert.match(typesSource, /previewItemId: string \| null/, '结果项缺少预览原文关联 ID')
assert.match(generationSource, /previewItemId:\s*item\.preview_item_id == null \? null : String\(item\.preview_item_id\)/, '结果映射丢失 preview_item_id')
assert.match(resultEditorSource, /previewItems: PreviewItem\[\]/, '结果编辑器缺少原文列表契约')
assert.match(resultEditorSource, /props\.previewItems\.find\(\(item\) => item\.id === previewItemId\)/, '结果编辑器没有按 preview_item_id 关联原文')
assert.match(resultEditorSource, /editedContent\.trim\(\)[\s\S]*?originalContent\.trim\(\)/, '原文参照没有优先展示实际用于生成的预处理内容')
assert.match(resultEditorSource, /原文参照/, '结果编辑器缺少原文参照区域')
assert.match(resultEditorSource, /实际送入模型的预处理后原文/, '结果编辑器没有说明原文参照口径')
assert.match(resultEditorSource, /selectedItem\.savedStatus === 'invalid'[\s\S]*?重新生成/, '失败结果右上角缺少单条重新生成按钮')
assert.match(resultEditorSource, /regeneratingResultId: string \| null/, '结果编辑器缺少单条重新生成 loading 契约')
assert.match(resultEditorSource, /emit\('regenerate:item', selectedItem\.id\)/, '失败结果按钮没有触发单条重新生成事件')
assert.match(viewSource, /:regenerating-result-id="regeneratingResultId"[\s\S]*?@regenerate:item="regenerateResult"/, '创建向导没有接入单条重新生成状态与事件')
assert.match(apiSource, /regenerateDataProcessResult[\s\S]*?\/results\/\$\{encodeURIComponent\(resultId\)\}\/regenerate/, '前端 API 缺少安全编码的单条重新生成接口')
assert.match(generationSource, /async function regenerateResult\(id: string\)[\s\S]*?regenerateDataProcessResult\(taskId, id,[\s\S]*?results\.value\[index\] = mapResult\(regenerated\)[\s\S]*?finally[\s\S]*?regeneratingResultId\.value = null/, '单条重新生成没有原位替换结果或可靠释放 loading')
assert.match(resultEditorSource, /invalidCount[\s\S]*?全部重新生成/, '结果编辑器缺少失败项一键全部重新生成入口')
assert.match(resultEditorSource, /emit\('regenerate:all'\)/, '全部重新生成按钮没有触发批量事件')
assert.match(resultEditorSource, /已处理[\s\S]*?成功[\s\S]*?失败/, '批量重新生成缺少总体进度与成功失败统计')
assert.doesNotMatch(resultEditorSource, /selectedItem\.qualityFlags|class="quality-flags"|v-for="flag in selectedItem\.qualityFlags"/, '结果编辑器不应直接渲染后端原始质量标签')
assert.doesNotMatch(typesSource, /qualityFlags/, '结果编辑器数据类型不应继续保留已移除的原始质量标签')
assert.doesNotMatch(generationSource, /qualityFlags:/, '结果映射不应继续创建已移除的原始质量标签')
assert.match(resultEditorSource, /selectedItem\.error[\s\S]*?validation-error/, '移除原始质量标签后仍应展示可读的失败原因')
assert.match(viewSource, /:bulk-regeneration="bulkRegeneration"[\s\S]*?@regenerate:all="regenerateAllResults"/, '创建向导没有接入全部重新生成状态与事件')
assert.match(apiSource, /regenerateDataProcessResults[\s\S]*?\/results\/regenerate-batch/, '前端 API 缺少批量重新生成接口')
assert.match(generationSource, /const BULK_REGENERATION_CHUNK_SIZE = 12/, '批量重新生成没有在接口超时预算内持续填满并发槽位')
assert.match(generationSource, /for \(let offset = 0; offset < candidates\.length; offset \+= BULK_REGENERATION_CHUNK_SIZE\)[\s\S]*?candidates\.slice\(offset, offset \+ BULK_REGENERATION_CHUNK_SIZE\)/, '批量重新生成循环没有实际使用受控批次大小')
assert.match(generationSource, /async function regenerateAllResults\(\)[\s\S]*?savedStatus === 'invalid'[\s\S]*?regenerateDataProcessResults[\s\S]*?bulkRegeneration\.completed/, '一键全部重新生成没有按服务端失败状态推进总体进度')
assert.match(generationSource, /bulkRegeneration\.status = 'partial'[\s\S]*?bulkRegeneration\.status = 'failed'/, '批量重新生成缺少部分成功与全部失败状态')
assert.match(typesSource, /savedStatus: 'valid' \| 'modified' \| 'invalid'/, '结果项缺少服务端保存状态,可能把未保存的本地错误误判为可重生成')
assert.match(viewSource, /watch\(processType,[\s\S]*?resetSourceDataForProcessTypeChange\(\)/, '切换处理类型后没有失效旧源数据')
assert.match(viewSource, /function resetSourceDataForProcessTypeChange\(\)[\s\S]*?uploadedFiles\.value = \[\][\s\S]*?selectedPreviewFileId\.value = null/, '旧源数据失效没有同步清理文件与预览选择')
assert.match(taskSetupSource, /v-if="processType === 'structured'"/, '结构化配置必须仅在结构化数据类型下显示')
const expectedStructuredGroups = [
[
"values: ['clean_invalid', 'deduplicate']",
'数据清洗',
'清理全空列和空记录,并删除内容完全相同的记录;不会猜测可空字段是否必填',
],
[
"values: ['detect_structure', 'normalize_format']",
'结构标准化',
'展平嵌套对象和可解析的 JSON 字段,并统一编码、空白、字段名和 JSON 序列化格式',
],
["values: ['desensitize']", '敏感信息脱敏', '识别并脱敏姓名、手机号、邮箱和身份证号'],
]
for (const [values, label, description] of expectedStructuredGroups) {
assert.ok(structuredOptionsSource.includes(values), `结构化预处理组合值不准确:${label}`)
assert.ok(structuredOptionsSource.includes(`label: '${label}'`), `结构化预处理缺少标签:${label}`)
assert.ok(structuredOptionsSource.includes(`description: '${description}'`), `结构化预处理语义不准确:${label}`)
}
assert.equal(expectedStructuredGroups.length, 3, '结构化预处理应收敛为 3 项')
const preprocessGroupsSource = structuredOptionsSource.slice(
structuredOptionsSource.indexOf('const PREPROCESS_GROUPS'),
structuredOptionsSource.indexOf('const legacyAnomalyFilterEnabled'),
)
assert.doesNotMatch(preprocessGroupsSource, /异常数据过滤|filter_anomaly|IQR/, '结构化新任务仍暴露异常数据过滤')
assert.match(structuredOptionsSource, /:indeterminate="groupIndeterminate\(group\.values\)"/, '历史部分选中的组合项没有半选回显')
assert.match(structuredOptionsSource, /function updatePreprocessGroup\([\s\S]*?new Set\(props\.options\.preprocessOptions\)[\s\S]*?next\.add\(value\)[\s\S]*?next\.delete\(value\)[\s\S]*?\[\.\.\.next\]/, '结构化预处理组合开关没有原子化更新或去重内部选项')
assert.match(typesSource, /仅用于恢复历史任务[\s\S]*?\| 'filter_anomaly'/, '异常数据过滤缺少历史兼容类型')
assert.match(structuredOptionsSource, /legacyAnomalyFilterEnabled[\s\S]*?历史任务[\s\S]*?结果可复现/, '历史异常过滤配置没有透明提示')
assert.ok(structuredOptionsSource.includes('生成选项'), '结构化配置缺少生成选项分类')
for (const splitName of ['训练集', '验证集', '测试集']) {
assert.ok(datasetSplitEditorSource.includes(splitName), `生成选项缺少数据集划分:${splitName}`)
}
assert.match(datasetSplitEditorSource, /const splitTotal = computed/, '数据集划分缺少比例总和计算')
assert.match(taskSetupSource, /splitTotal\.value !== 100/, '数据集划分缺少总和 100% 校验')
assert.ok(datasetSplitEditorSource.includes('训练集、验证集和测试集比例总和必须为 100%'), '数据集划分缺少就地错误提示')
for (const splitField of ['train', 'validation', 'test']) {
assert.match(
datasetSplitEditorSource,
new RegExp(`modelValue\\.${splitField}[\\s\\S]*?:min="0"[\\s\\S]*?:max="100"[\\s\\S]*?:step="1"[\\s\\S]*?:precision="0"`),
`数据集划分字段 ${splitField} 缺少 0100 的整数限制`,
)
}
assert.match(typesSource, /QA_PAIRS_GENERATION_LIMITS\s*=\s*\{ min: 1, max: 50 \}/, '问答生成数量统一范围必须为 1 到 50')
assert.match(typesSource, /function normalizeQaPairsGenerationCount[\s\S]*?Math\.trunc\(parsed\)[\s\S]*?QA_PAIRS_GENERATION_LIMITS\.max[\s\S]*?QA_PAIRS_GENERATION_LIMITS\.min/, '问答生成数量缺少统一整数归一化和边界裁剪')
assert.match(structuredOptionsSource, /options\.qaPairsPerRow[\s\S]*?:min="QA_PAIRS_GENERATION_LIMITS\.min"[\s\S]*?:max="QA_PAIRS_GENERATION_LIMITS\.max"/, '每行生成数量必须使用统一的 1 到 50 限制')
assert.match(structuredOptionsSource, /支持 150 条;数量越大,处理耗时和 Token 消耗越高/, '结构化生成数量缺少耗时与 Token 消耗说明')
assert.match(viewSource, /const structuredOptions = ref<StructuredProcessOptions>/, '父页面缺少结构化配置状态')
assert.match(stateSource, /datasetSplit:\s*\{ train: 80, validation: 10, test: 10 \}/, '数据集划分默认值必须为 80/10/10')
assert.match(viewSource, /v-model:structured-options="structuredOptions"/, '父页面没有双向绑定结构化配置')
assert.match(generationSource, /generateDataProcess\(taskId\)/, '开始生成没有调用真实 API')
assert.match(generationSource, /getDataProcessProgress\(taskId\)/, '生成状态没有通过真实 API 轮询')
assert.match(generationSource, /getDataProcessResults\(taskId,[\s\S]*?page:[\s\S]*?page_size:/, '生成完成后没有分页加载真实结果')
assert.match(generationSource, /updateDataProcessResult\(taskId,\s*item\.id,[\s\S]*?expected_updated_at:/, '结果保存没有调用真实 API 或缺少并发版本')
assert.ok(generationSource.includes('item.quality_score?.overall'), '结果映射没有读取质量总分 overall')
assert.doesNotMatch(generationSource, /createResults\(/, '生成 composable 仍在本地伪造处理结果')
for (const apiName of [
'createDataProcessTask',
'updateDataProcessWorkflowStep',
'regenerateDataProcessTask',
'uploadDataProcessSourceFiles',
'buildDataProcessPreview',
'startDataProcessPreview',
'getDataProcessPreviewProgress',
'getDataProcessPreview',
'generateDataProcess',
'getDataProcessProgress',
'getDataProcessResults',
'updateDataProcessResult',
'confirmDataProcessResults',
'publishDataProcess',
]) {
assert.match(apiSource, new RegExp(`export (?:const|async function|function) ${apiName}\\b`), `API 模块缺少 ${apiName}`)
}
assert.match(sourceUploadWorkerSource, /uploadDataProcessSourceFiles\(currentTaskId,\s*\[job\.file\]/, '文件上传没有逐文件调用真实 API')
assert.match(sourceUploadWorkerSource, /STRUCTURED_FILE_EXTENSIONS = new Set\(\['json', 'jsonl', 'ndjson', 'csv', 'tsv', 'xlsx'\]\)/, '结构化文件扩展名白名单不完整')
for (const extension of ['txt', 'md', 'markdown', 'pdf', 'docx', 'pptx', 'json', 'jsonl', 'ndjson']) {
assert.ok(sourceUploadWorkerSource.includes(`'${extension}'`), `非结构化文件扩展名白名单缺少 ${extension}`)
}
assert.match(sourceUploadWorkerSource, /LEGACY_OFFICE_EXTENSIONS = new Set\(\['doc', 'xls', 'ppt'\]\)/, '缺少旧版 Office 格式识别')
assert.ok(sourceUploadWorkerSource.includes('请分别转换为 DOCX、XLSX、PPTX 后上传'), '旧版 Office 文件缺少转换提示')
const sourceValidationStart = sourceUploadWorkerSource.indexOf('export function validateSourceFileSelection(')
const sourceValidationEnd = sourceUploadWorkerSource.indexOf('\n}\n\nfunction unicodeCodePointLength', sourceValidationStart)
assert.ok(sourceValidationStart >= 0 && sourceValidationEnd > sourceValidationStart, '无法定位源文件选择校验函数')
const sourceValidationSource = sourceUploadWorkerSource.slice(sourceValidationStart, sourceValidationEnd + 2)
assert.doesNotMatch(sourceValidationSource, /file\.name === raw\.name[\s\S]{0,160}file\.size === raw\.size|同名且同大小/, '不同内容但同名同大小的文件仍会被前端误拒绝')
assert.match(sourceValidationSource, /selectedFiles\.length >= MAX_SOURCE_FILE_COUNT/, '移除伪重复校验时误删了文件数量限制')
assert.match(sourceValidationSource, /selectedBytes \+ raw\.size > MAX_SOURCE_BATCH_BYTES/, '移除伪重复校验时误删了批次大小限制')
assert.doesNotMatch(sourceUploadWorkerSource, /job\.file\.arrayBuffer\(|new TextDecoder/, '上传前仍把整个文本文件读入浏览器内存')
assert.match(sourceUploadWorkerSource, /export async function loadCanonicalSourceContent[\s\S]*?offset,[\s\S]*?limit: SOURCE_CONTENT_PAGE_CHARS/, '服务端 canonical content 没有按有界字符窗口读取')
assert.match(sourceUploadWorkerSource, /pending\.content = await loadCanonicalSourceContent\(currentTaskId, source\.id\)/, '上传成功后没有统一使用服务端 canonical content')
assert.doesNotMatch(sourceUploadWorkerSource, /\brawFile:\s*job\.file\b/, '上传成功状态仍长期保留原始 File')
assert.doesNotMatch(typesSource, /\brawFile\??:\s*File\b/, '上传状态类型仍长期持有原始 File')
assert.doesNotMatch(viewSource, /\brawFile:\s*raw\b/, '待上传列表仍复制保存原始 File')
assert.match(apiSource, /params:\s*\{[\s\S]*?offset\?: number[\s\S]*?limit\?: number[\s\S]*?\}/, '正文 API 前端契约缺少字符窗口参数')
assert.match(apiSource, /formData\.append\('files', file\)/, '上传 API 没有使用 files 多文件表单字段')
assert.match(apiSource, /onUploadProgress:[\s\S]*?event\.loaded \/ event\.total[\s\S]*?Math\.min\(99,/, '上传 API 没有接入真实字节进度或响应前未限制在 99%')
assert.match(apiSource, /source-files`[\s\S]*?timeout: 5 \* 60 \* 1000/, '源文件上传缺少 5 分钟超时')
assert.match(apiSource, /\/preview\/build/, 'API 模块缺少后端预览构建路径')
assert.match(apiSource, /\/preview\/build`[\s\S]*?\{ timeout: 5 \* 60 \* 1000 \}/, '单文件切分请求缺少 5 分钟超时')
assert.match(apiSource, /\/preview\/start`/, 'API 模块缺少后台切分启动路径')
assert.match(apiSource, /\/preview\/progress`/, 'API 模块缺少后台切分进度路径')
assert.match(apiSource, /\/workflow-step`/, 'API 模块缺少向导步骤持久化路径')
assert.match(apiSource, /\/progress`/, 'API 模块缺少生成进度路径')
assert.match(apiSource, /\/results`/, 'API 模块缺少结果分页路径')
assert.match(apiSource, /\/publish`/, 'API 模块缺少数据集发布路径')
assert.match(contractTypesSource, /source_file_ids\?: Array<string \| number>/, '预览构建契约缺少源文件 ID 列表')
assert.match(
contractTypesSource,
/export type DataProcessPreviewFileStatus = 'waiting' \| 'processing' \| 'success' \| 'failed'/,
'文件预览状态契约不完整',
)
assert.match(
contractTypesSource,
/export interface DataProcessPreviewProgress[\s\S]*?workflow_step: DataProcessWorkflowStep[\s\S]*?preview_status: DataProcessPreviewStatus[\s\S]*?preview_progress: number[\s\S]*?preview_run_id/,
'后台切分进度契约缺少步骤、状态、进度或任务代次',
)
for (const field of ['status', 'uploadProgress', 'uploadError', 'previewStatus', 'previewProgress', 'previewError', 'previewConfigSignature']) {
assert.ok(typesSource.includes(field), `上传文件缺少逐文件预览字段:${field}`)
}
assert.match(typesSource, /status: 'queued' \| 'uploading' \| 'ready' \| 'failed'/, '上传文件状态机不完整')
assert.match(viewSource, /uploadedFiles\.value\.push\([\s\S]*?status: 'queued'[\s\S]*?enqueueSourceUpload/, '文件选择后没有先进入列表再加入上传队列')
assert.match(sourceUploadWorkerSource, /while \(queue\.length\) \{[\s\S]*?await uploadOne\(job\)/, '多文件上传没有由单一队列逐个等待')
assert.doesNotMatch(sourceUploadWorkerSource, /Promise\.(?:all|allSettled)/, '上传队列不得并发消费文件')
assert.match(sourceUploadWorkerSource, /pending\.status = 'ready'[\s\S]*?pending\.uploadProgress = 100/, '服务端响应成功后没有将文件置为上传完成')
assert.match(nextFromUploadSource, /failedUploads[\s\S]*?hasUnfinishedUploads[\s\S]*?monitorPreviewBuild\(pendingFileIds\)/, '上传失败或未完成时没有阻断后台切分')
assert.doesNotMatch(viewSource, /file\.sourceFileId \|\| file\.uid/, '切分或删除仍可能把本地临时 UID 当成后端文件 ID')
assert.match(contractTypesSource, /expected_updated_at\?: string/, '编辑契约缺少乐观并发版本字段')
assert.match(contractTypesSource, /export interface DataProcessRegeneratePayload extends DataProcessTaskCreatePayload[\s\S]*?expected_updated_at: string/, '重新生成 payload 的原任务并发版本必须为必填')
assert.match(contractTypesSource, /export interface DataProcessRegenerateResult[\s\S]*?preview_invalidated: boolean[\s\S]*?published_outputs_preserved: boolean/, '重新生成响应缺少预览失效或已发布数据保留状态')
assert.match(apiSource, /regenerateDataProcessTask[\s\S]*?\/regenerate`/, '重新生成 API 路径未接入')
for (const field of [
'generationModelId',
'generationPrompt',
'outputType',
'reasoningDetail',
'qualityFilterEnabled',
'filterLowQuality',
'filterShortContent',
'minOutputLength',
]) {
assert.ok(typesSource.includes(field), `生成控制配置缺少字段:${field}`)
assert.ok(implementationSource.includes(field), `父页面默认值缺少字段:${field}`)
}
assert.match(structuredOptionsSource, /GenerationOptionsPanel/, '结构化生成选项没有复用统一的质量筛选组件')
assert.match(unstructuredOptionsSource, /GenerationOptionsPanel/, '非结构化生成选项没有复用统一的质量筛选组件')
assert.match(structuredOptionsSource, /:options="options"/, '结构化生成选项未接入统一配置组件')
assert.match(unstructuredOptionsSource, /:options="options"/, '非结构化生成选项未接入统一配置组件')
assert.doesNotMatch(taskSetupFeatureSource, /<h3>大模型<\/h3>|section="model"/, '第一步不应继续承载大模型配置')
assert.match(modelSelectionSource, /<h3[^>]*>大模型选择<\/h3>/, '独立步骤缺少大模型选择标题')
assert.match(modelSelectionSource, /section="model"/, '独立步骤没有挂载模型配置')
assert.match(modelSelectionSource, /defineExpose\(\{ validate \}\)/, '独立大模型选择步骤没有暴露继续前校验')
assert.match(modelSelectionSource, /class="form-section"/, '大模型选择步骤没有沿用第一步的通栏表单分区')
assert.doesNotMatch(modelSelectionSource, /max-width:\s*980px/, '大模型选择步骤不应使用比第一步更窄的固定内容宽度')
assert.match(taskSetupFeatureSource, /section="quality"/, '质量筛选没有保留在生成选项分类中')
assert.doesNotMatch(generationControlSource, /<h4>大模型<\/h4>/, '大模型不应继续作为生成选项内部子分类')
for (const label of ['大模型', '数据生成模型', '默认提示语', '输出类型', '标准回答', '思维链回答', '推理详细程度', '普通推理(推荐)', '详细推理', '质量筛选', '过滤低质量内容', '过滤过短内容', '最少字数']) {
assert.ok(generationControlSource.includes(label), `生成控制界面缺少:${label}`)
}
assert.match(generationControlSource, /filterable/, '数据生成模型下拉必须支持搜索')
assert.match(generationControlSource, /maxlength="2000"/, '默认提示语缺少合理的长度限制')
assert.match(generationControlSource, /aria-label="输出类型"/, '输出类型选项缺少可访问名称')
assert.match(generationControlSource, /v-if="options\.outputType === 'reasoning'" class="output-type-row"/, '推理详细程度没有按思维链模式渐进显示')
assert.match(generationControlSource, /aria-label="推理详细程度"/, '推理详细程度缺少可访问名称')
assert.match(generationControlSource, /&lt;think&gt;推理过程&lt;\/think&gt;/, '思维链选项没有说明最终保存格式')
assert.match(generationControlSource, /<el-select[\s\S]*?class="output-type-select"[\s\S]*?aria-label="输出类型"/, '输出类型必须使用右侧下拉选择')
assert.doesNotMatch(generationControlSource, /class="output-type-options"/, '输出类型不应继续使用横向按钮组')
assert.match(generationControlSource, /\.output-type-row\s*\{[\s\S]*?grid-template-columns:\s*minmax\(0, 1fr\) 150px/, '输出类型没有与生成数量控件保持一致宽度')
assert.match(generationControlSource, /\.output-type-select\s*\{[\s\S]*?width:\s*150px/, '输出类型下拉框宽度没有与生成数量控件对齐')
assert.match(
generationControlSource,
/<div v-else class="generation-config-group quality-config-group">[\s\S]*?<strong>输出类型<\/strong>[\s\S]*?<div class="quality-switch-row">/,
'输出类型必须显示在第一步生成选项中,并位于质量筛选之前',
)
const modelGenerationSection = generationControlSource.slice(
generationControlSource.indexOf('<div v-if="section === \'model\'"'),
generationControlSource.indexOf('<div v-else class="generation-config-group quality-config-group">'),
)
assert.doesNotMatch(modelGenerationSection, /<strong>输出类型<\/strong>/, '第二步大模型选择不应重复显示输出类型')
assert.match(generationControlSource, /\.model-field\s*\{[\s\S]*?display:\s*flex[\s\S]*?flex-direction:\s*column/, '大模型字段没有使用稳定的纵向表单布局')
assert.match(generationControlSource, /\.generation-config-group\s*\{[\s\S]*?border:\s*1px solid #e2e5ec/, '大模型配置没有保留统一配置面板边框')
assert.match(generationControlSource, /\.model-config-group\s*\{[\s\S]*?padding:\s*0[\s\S]*?border:\s*0/, '独立大模型步骤仍存在嵌套卡片挤压')
assert.match(
generationControlSource,
/\.model-config-group \.advanced-settings-grid\s*\{[\s\S]*?grid-template-columns:\s*1fr/,
'大模型高级参数没有改为与第一步一致的纵向布局',
)
function defaultPromptValue(name) {
const match = stateSource.match(new RegExp(`export const ${name} = ` + '`([\\s\\S]*?)`'))
assert.ok(match, `${name} 缺少可编辑的内置默认提示语`)
return match[1]
}
const standardPrompt = defaultPromptValue('DEFAULT_STANDARD_GENERATION_PROMPT')
const reasoningPrompt = defaultPromptValue('DEFAULT_REASONING_GENERATION_PROMPT')
assert.ok(standardPrompt.includes('{{ content }}'), '标准回答默认提示语缺少 {{ content }} 占位符')
assert.ok(reasoningPrompt.includes('{{ content }}'), '思维链默认提示语缺少 {{ content }} 占位符')
assert.ok(standardPrompt.length <= 2000, '标准回答默认提示语超过输入框长度限制')
assert.ok(reasoningPrompt.length <= 2000, '思维链默认提示语超过输入框长度限制')
assert.ok(reasoningPrompt.includes('逐步展开思考过程'), '思维链默认提示语没有明确要求逐步生成思考过程')
assert.ok(reasoningPrompt.includes('不得跳过关键步骤只给结论'), '思维链默认提示语没有禁止只生成最终答案')
assert.notEqual(standardPrompt, reasoningPrompt, '标准回答与思维链回答不得共用默认提示语')
assert.match(generationControlSource, /默认提示语已包含[\s\S]*?\{\{ content \}\}[\s\S]*?可移动该占位符/, '默认提示语没有说明 {{ content }} 的内置位置和修改方式')
assert.equal((stateSource.match(/generationPrompt:\s*DEFAULT_STANDARD_GENERATION_PROMPT/g) || []).length, 2, '结构化与非结构化任务应默认使用标准回答提示语')
assert.match(generationControlSource, /isBuiltInGenerationPrompt\(props\.options\.generationPrompt\)[\s\S]*?defaultGenerationPrompt\(outputType\)[\s\S]*?: props\.options\.generationPrompt/, '切换输出类型时没有在保留自定义提示语的前提下切换内置提示语')
assert.equal((stateSource.match(/^\s{4}outputType:\s*'standard',/gm) || []).length, 2, '结构化与非结构化任务应默认生成标准回答')
assert.equal((stateSource.match(/reasoningDetail:\s*'normal'/g) || []).length, 2, '结构化与非结构化任务应默认使用普通推理')
assert.match(generationControlSource, /v-if="options\.qualityFilterEnabled"/, '质量规则没有随总开关渐进显示')
assert.match(generationControlSource, /v-if="options\.filterShortContent"/, '最少字数没有随短内容规则显示')
assert.match(generationControlSource, /:min="1"[\s\S]*:max="1000"/, '最少字数缺少 1 到 1000 的边界限制')
assert.match(taskSetupSource, /qualityValidationMessage/, '质量规则缺少继续前校验')
assert.match(viewSource, /useModelsStore/, '创建页没有加载模型列表')
const generationModelsStart = viewSource.indexOf('const generationModels')
const generationModelsEnd = viewSource.indexOf('const taskSetupRef', generationModelsStart)
const generationModelsSource = viewSource.slice(generationModelsStart, generationModelsEnd)
assert.match(generationModelsSource, /computed\(\(\) => modelList\.value\)/, '数据生成模型候选没有直接使用模型管理完整列表')
assert.doesNotMatch(generationModelsSource, /\.filter\(|model\.type|model\.model_source|model\.status|model\.purpose/, '数据生成模型候选仍按类型、来源、状态或用途静默过滤')
assert.match(viewSource, /modelsStore\.load\(true\)/, '进入创建向导时没有强制刷新模型管理列表')
assert.match(viewSource, /<ModelSelectionStep[\s\S]*?:models="generationModels"/, '创建页没有向独立大模型选择步骤传递模型列表')
assert.match(generationControlSource, /v-for="model in models \|\| \[\]"/, '模型下拉没有遍历完整候选列表')
assert.match(generationControlSource, /:value="model\.id"/, '模型下拉没有使用模型管理 ID 作为选中值')
assert.match(generationControlSource, /modelMeta\(model\)/, '模型下拉缺少来源和类型说明')
assert.match(typesSource, /export interface UnstructuredProcessOptions/, '缺少非结构化处理选项类型')
assert.match(
typesSource,
/export type ChunkMethod = 'layout_hybrid' \| 'semantic' \| 'fixed'/,
'非结构化切分方式类型必须只保留 layout_hybrid、semantic 和 fixed',
)
for (const field of [
'preprocessOptions',
'chunkMethod',
'chunkSize',
'chunkOverlap',
'minChunkSize',
'semanticBreakpointPercentile',
'preserveTables',
'preserveCodeBlocks',
'preserveLists',
'semanticEnrichment',
'qaPairsPerChunk',
'datasetSplit',
]) {
assert.ok(typesSource.includes(field), `非结构化处理选项缺少字段:${field}`)
}
for (const removedField of ['contextScope', 'generationTypes', 'skipUnanswerable']) {
assert.ok(!typesSource.includes(removedField), `简化后仍保留低频生成字段:${removedField}`)
}
assert.match(taskSetupSource, /v-if="processType === 'unstructured'"/, '非结构化配置必须仅在非结构化数据类型下显示')
assert.match(taskSetupSource, /processTypeLocked\?: boolean/, '任务配置缺少处理类型锁定状态')
assert.match(taskSetupSource, /:disabled="processTypeLocked"/, '重新生成时处理类型选项没有禁用')
assert.ok(taskSetupSource.includes('重新生成沿用原任务处理类型,不可修改'), '处理类型锁定缺少明确说明')
assert.match(viewSource, /:process-type-locked="isRegeneration"/, '创建向导没有向第一步传递重新生成锁定状态')
assert.ok(unstructuredOptionsSource.includes('智能预处理'), '简化后缺少智能预处理总开关')
assert.ok(unstructuredOptionsSource.includes('敏感信息脱敏'), '简化后缺少脱敏开关')
const expectedSmartPreprocessOptions = [
'clean_invalid_content',
'detect_document_structure',
'merge_short_content',
'filter_low_quality',
'deduplicate_content',
'preserve_context',
]
const smartOptionsStart = unstructuredOptionsSource.indexOf('const SMART_PREPROCESS_OPTIONS')
const smartOptionsEnd = unstructuredOptionsSource.indexOf('const CHUNK_METHODS', smartOptionsStart)
const smartOptionsSource = unstructuredOptionsSource.slice(smartOptionsStart, smartOptionsEnd)
const smartOptionValues = [...smartOptionsSource.matchAll(/^\s*'([^']+)',?$/gm)].map((match) => match[1])
assert.deepEqual(smartOptionValues, expectedSmartPreprocessOptions, '智能预处理内部值与后端语义不一致')
assert.equal(new Set(smartOptionValues).size, smartOptionValues.length, '非结构化智能预处理 value 必须唯一')
for (const descriptionPart of [
'清理页眉页脚、页码、目录和无效内容',
'感知文档结构',
'合并短块',
'预过滤低质量内容',
'近重复去重',
'重叠保护上下文',
]) {
assert.ok(unstructuredOptionsSource.includes(descriptionPart), `智能预处理说明缺少语义:${descriptionPart}`)
}
assert.ok(unstructuredOptionsSource.includes('姓名、手机号、邮箱和身份证号'), '非结构化脱敏说明缺少完整字段范围')
assert.match(unstructuredOptionsSource, /Array\.from\(new Set\(props\.options\.preprocessOptions\.filter\(/, '非结构化预处理选中值没有去重')
assert.match(unstructuredOptionsSource, /const smartPreprocessEnabled = computed/, '智能预处理没有映射到内部处理项')
assert.match(unstructuredOptionsSource, /function updateSmartPreprocess/, '智能预处理开关缺少更新逻辑')
assert.match(unstructuredOptionsSource, /function updateDesensitize/, '脱敏开关缺少更新逻辑')
assert.ok(unstructuredOptionsSource.includes('切分选项'), '非结构化配置缺少切分选项分类')
const expectedChunkMethods = [
['layout_hybrid', '版面结构混合切分'],
['semantic', '语义切分'],
['fixed', '固定 Token 切分'],
]
const chunkMethodsStart = unstructuredOptionsSource.indexOf('const CHUNK_METHODS')
const chunkMethodsEnd = unstructuredOptionsSource.indexOf('const UNSTRUCTURED_NUMBER_LIMITS', chunkMethodsStart)
const chunkMethodsSource = unstructuredOptionsSource.slice(chunkMethodsStart, chunkMethodsEnd)
const chunkMethodValues = [...chunkMethodsSource.matchAll(/value: '([^']+)'/g)].map((match) => match[1])
assert.deepEqual(chunkMethodValues, expectedChunkMethods.map(([value]) => value), '切分方式值集合不准确')
for (const [value, label] of expectedChunkMethods) {
assert.ok(chunkMethodsSource.includes(`value: '${value}', label: '${label}'`), `切分方式缺少选项:${label}`)
}
for (const removedMethod of ['structure', 'custom', 'heading']) {
assert.ok(!chunkMethodsSource.includes(`value: '${removedMethod}'`), `切分方式仍保留已移除值:${removedMethod}`)
}
assert.ok(unstructuredOptionsSource.includes('默认推荐,按标题、段落、列表和表格结构切分'), '版面结构混合切分缺少推荐说明')
assert.ok(unstructuredOptionsSource.includes('根据相邻内容的语义变化寻找主题边界'), '语义切分缺少适用场景说明')
assert.ok(unstructuredOptionsSource.includes('按句子边界控制固定 Token 长度'), '固定 Token 切分缺少稳定性说明')
assert.doesNotMatch(unstructuredOptionsSource, /自定义分隔符/, '切分选项仍保留自定义分隔符')
assert.doesNotMatch(taskSetupSource, /customDelimiter|自定义分隔符/, '任务配置仍校验已移除的自定义分隔符')
for (const label of ['切片长度', '重叠长度', '最小切片长度', '保护表格、代码和列表']) {
assert.ok(unstructuredOptionsSource.includes(label), `切分选项缺少配置:${label}`)
}
assert.doesNotMatch(unstructuredOptionsSource, /advancedChunkSettingsOpen|>高级设置</, '切分核心参数不应再隐藏在高级设置中')
assert.match(taskSetupSource, /if \(chunkValidationMessage\.value\) \{[\s\S]*?revealValidation\(\)[\s\S]*?return false/, '高级切分配置校验失败时没有重新展开定位')
assert.match(unstructuredOptionsSource, /const preserveSpecialContentEnabled = computed/, '特殊内容保护没有合并为单一开关')
assert.match(unstructuredOptionsSource, /function updateSpecialContentProtection/, '特殊内容保护开关缺少更新逻辑')
assert.match(unstructuredOptionsSource, /options\.chunkSize[\s\S]*?:min="200"[\s\S]*?:max="2000"/, '切片长度必须限制在 200 到 2000 Token')
assert.match(unstructuredOptionsSource, /options\.chunkOverlap[\s\S]*?:min="0"[\s\S]*?:max="500"/, '重叠长度必须限制在 0 到 500 Token')
assert.match(unstructuredOptionsSource, /options\.minChunkSize[\s\S]*?:min="20"[\s\S]*?:max="500"/, '最小切片长度必须限制在 20 到 500 Token')
assert.match(unstructuredOptionsSource, /options\.semanticBreakpointPercentile[\s\S]*?:min="1"[\s\S]*?:max="99"/, '语义断点百分位必须限制在 1 到 99')
assert.match(unstructuredOptionsSource, /v-if="options\.chunkMethod === 'semantic'"[\s\S]*?options\.semanticBreakpointPercentile/, '语义断点百分位必须仅在语义切分时显示')
for (const label of ['每个切片生成数量', '数据集划分']) {
assert.ok(taskSetupFeatureSource.includes(label), `非结构化生成选项缺少:${label}`)
}
for (const removedLabel of ['上下文范围', '问题类型', '跳过无法回答的内容']) {
assert.ok(!taskSetupFeatureSource.includes(removedLabel), `简化后仍显示低频选项:${removedLabel}`)
}
assert.match(unstructuredOptionsSource, /qaPairsPerChunk:\s*QA_PAIRS_GENERATION_LIMITS/, '每切片生成数量裁剪必须使用统一的 1 到 50 限制')
assert.match(unstructuredOptionsSource, /options\.qaPairsPerChunk[\s\S]*?:min="QA_PAIRS_GENERATION_LIMITS\.min"[\s\S]*?:max="QA_PAIRS_GENERATION_LIMITS\.max"/, '每个切片生成数量必须使用统一的 1 到 50 限制')
assert.match(unstructuredOptionsSource, /支持 150 条;数量越大,处理耗时和 Token 消耗越高/, '非结构化生成数量缺少耗时与 Token 消耗说明')
assert.match(taskSetupSource, /unstructuredSplitTotal\.value !== 100/, '非结构化数据集划分缺少总和 100% 校验')
assert.match(taskSetupSource, /chunkOverlap \+ props\.unstructuredOptions\.minChunkSize[\s\S]*?> props\.unstructuredOptions\.chunkSize/, '切分配置未校验重叠长度与最小切片长度的组合边界')
assert.ok(unstructuredOptionsSource.includes('Token 数为轻量估算值'), '切片长度缺少 Token 估算说明')
assert.match(unstructuredOptionsSource, /\.chunk-settings-grid\s*\{[\s\S]*?grid-template-columns:\s*repeat\(3,\s*minmax\(0,\s*1fr\)\)/, '核心切分参数没有收紧为三列布局')
assert.match(viewSource, /const unstructuredOptions = ref<UnstructuredProcessOptions>/, '父页面缺少非结构化配置状态')
assert.match(stateSource, /chunkMethod:\s*'layout_hybrid'/, '非结构化默认切分方式必须为版面结构混合切分')
assert.match(stateSource, /chunkSize:\s*800/, '默认切片长度必须为 800 Token')
assert.match(stateSource, /chunkOverlap:\s*100/, '默认重叠长度必须为 100 Token')
assert.match(stateSource, /minChunkSize:\s*100/, '默认最小切片长度必须为 100 Token')
assert.match(stateSource, /semanticBreakpointPercentile:\s*95/, '默认语义断点百分位必须为 95')
assert.match(stateSource, /qaPairsPerChunk:\s*1/, '默认每个切片必须生成 1 个问答对')
assert.match(stateSource, /qaPairsPerRow:\s*1/, '默认每行必须生成 1 个问答对')
assert.match(viewSource, /v-model:unstructured-options="unstructuredOptions"/, '父页面没有双向绑定非结构化配置')
assert.match(viewSource, /JSON\.stringify\(previewAffectingOptions\(\)\)/, '影响切分的非结构化配置没有纳入预览失效判断')
function defaultPreprocessValues(functionName, nextFunctionName) {
const start = stateSource.indexOf(`export function ${functionName}`)
const end = nextFunctionName ? stateSource.indexOf(`export function ${nextFunctionName}`, start) : stateSource.length
const functionSource = stateSource.slice(start, end)
const match = functionSource.match(/preprocessOptions:\s*\[([\s\S]*?)\]/)
assert.ok(match, `${functionName} 缺少 preprocessOptions 默认值`)
return [...match[1].matchAll(/'([^']+)'/g)].map((item) => item[1])
}
const defaultStructuredPreprocess = defaultPreprocessValues(
'createDefaultStructuredOptions',
'createDefaultUnstructuredOptions',
)
assert.deepEqual(
defaultStructuredPreprocess,
[],
'结构化新任务不应默认勾选预处理',
)
assert.equal(new Set(defaultStructuredPreprocess).size, defaultStructuredPreprocess.length, '结构化默认预处理值重复')
const defaultUnstructuredPreprocess = defaultPreprocessValues('createDefaultUnstructuredOptions')
assert.deepEqual(defaultUnstructuredPreprocess, [], '非结构化新任务不应默认勾选预处理')
assert.equal(new Set(defaultUnstructuredPreprocess).size, defaultUnstructuredPreprocess.length, '非结构化默认预处理值重复')
for (const field of ['preserveTables', 'preserveCodeBlocks', 'preserveLists']) {
assert.match(
stateSource,
new RegExp(`${field}:\\s*false`),
`非结构化预处理选项 ${field} 不应默认开启`,
)
}
assert.match(structuredOptionsSource, /默认不执行预处理,请按数据情况自行选择/, '结构化预处理缺少默认不勾选说明')
assert.match(unstructuredOptionsSource, /默认不执行预处理,请按文档情况自行选择/, '非结构化预处理缺少默认不勾选说明')
assert.doesNotMatch(unstructuredOptionsSource, /默认启用结构感知/, '非结构化预处理仍保留默认启用的误导文案')
const backendConfigStart = viewSource.indexOf('function toBackendConfig()')
const backendConfigEnd = viewSource.indexOf('function taskPayload()', backendConfigStart)
assert.ok(backendConfigStart >= 0 && backendConfigEnd > backendConfigStart, '缺少任务后端配置映射')
const backendConfigSource = viewSource.slice(backendConfigStart, backendConfigEnd)
assert.ok(backendConfigSource.includes('preprocess_options: [...options.preprocessOptions]'), '预处理选项没有完整传入任务配置')
assert.ok(backendConfigSource.includes('dataset_split: { ...options.datasetSplit }'), '数据集划分没有完整传入任务配置')
for (const [backendField, frontendField] of [
['semantic_enrichment', 'semanticEnrichment'],
['generation_model_id', 'generationModelId'],
['generation_prompt', 'generationPrompt'],
['output_type', 'outputType'],
['reasoning_detail', 'reasoningDetail'],
['temperature', 'temperature'],
['max_tokens', 'maxTokens'],
['json_mode', 'jsonMode'],
['quality_filter_enabled', 'qualityFilterEnabled'],
['filter_low_quality', 'filterLowQuality'],
['filter_short_content', 'filterShortContent'],
['min_output_length', 'minOutputLength'],
]) {
assert.ok(
backendConfigSource.includes(`${backendField}: options.${frontendField}`),
`公共配置 ${frontendField} 没有传入 task payload`,
)
}
for (const [backendField, frontendField] of [
['chunk_method', 'chunkMethod'],
['chunk_size', 'chunkSize'],
['chunk_overlap', 'chunkOverlap'],
['min_chunk_size', 'minChunkSize'],
['semantic_breakpoint_percentile', 'semanticBreakpointPercentile'],
['preserve_tables', 'preserveTables'],
['preserve_code_blocks', 'preserveCodeBlocks'],
['preserve_lists', 'preserveLists'],
['qa_pairs_per_chunk', 'qaPairsPerChunk'],
]) {
assert.ok(
backendConfigSource.includes(`${backendField}: unstructuredOptions.value.${frontendField}`),
`非结构化配置 ${frontendField} 没有传入 task payload`,
)
}
assert.ok(
backendConfigSource.includes('qa_pairs_per_row: structuredOptions.value.qaPairsPerRow'),
'结构化每行生成数量没有传入 task payload',
)
const taskPayloadStart = viewSource.indexOf('function taskPayload()')
const taskPayloadEnd = viewSource.indexOf('function externalPayload()', taskPayloadStart)
const taskPayloadSource = viewSource.slice(taskPayloadStart, taskPayloadEnd)
for (const marker of ['name: task.name.trim()', 'description: task.description.trim()', 'process_type: originalProcessType.value || processType.value', 'config: toBackendConfig()']) {
assert.ok(taskPayloadSource.includes(marker), `任务创建 payload 缺少:${marker}`)
}
assert.match(stateSource, /Object\.prototype\.hasOwnProperty\.call\(config, key\)/, '配置反向映射没有区分缺失值与 false/0')
assert.match(stateSource, /Number\.isFinite\(value\) \? value : fallback/, '配置反向映射没有保留合法数字 0')
assert.match(stateSource, /qaPairsPerRow:\s*normalizeQaPairsGenerationCount\([\s\S]*?qa_pairs_per_row[\s\S]*?defaults\.qaPairsPerRow/, '结构化生成数量回填没有按 1 到 50 归一化')
assert.match(stateSource, /qaPairsPerChunk:\s*normalizeQaPairsGenerationCount\([\s\S]*?qa_pairs_per_chunk[\s\S]*?defaults\.qaPairsPerChunk/, '非结构化生成数量回填没有按 1 到 50 归一化')
assert.match(stateSource, /const outputType = configValue\(config, 'output_type', defaults\.outputType\) === 'reasoning'[\s\S]*?\? 'reasoning'[\s\S]*?: 'standard'/, '输出类型没有从任务配置安全回填')
assert.match(stateSource, /reasoningDetail:\s*configValue\(config, 'reasoning_detail', defaults\.reasoningDetail\) === 'detailed'[\s\S]*?\? 'detailed'[\s\S]*?: 'normal'/, '推理详细程度没有从任务配置安全回填')
assert.match(stateSource, /isBuiltInGenerationPrompt\(configuredPrompt\)[\s\S]*?defaultGenerationPrompt\(outputType\)/, '旧版内置提示语没有按输出类型迁移')
assert.match(stateSource, /createStructuredOptionsFromConfig/, '结构化配置缺少后端到表单的反向映射')
assert.match(stateSource, /createUnstructuredOptionsFromConfig/, '非结构化配置缺少后端到表单的反向映射')
assert.match(stateSource, /configValue<unknown>\(config, 'preprocess_options', \[\]\)/, '历史任务缺少预处理配置时必须按后端空列表语义回填')
for (const [field, fallback] of [
['preserve_tables', 'preserveTables'],
['preserve_code_blocks', 'preserveCodeBlocks'],
['preserve_lists', 'preserveLists'],
]) {
assert.match(
stateSource,
new RegExp(`configValue\\(\\s*config,\\s*'${field}',[\\s\\S]*?defaults\\.${fallback}`),
`历史任务缺少 ${field} 时必须沿用表单默认值,同时保留显式 false`,
)
}
assert.match(regenerationSource, /getDataProcessTask\(sourceTaskId\.value\)/, '重新生成没有加载原任务')
assert.match(regenerationSource, /loadCanonicalSourceContent\(taskId, file\.id\)/, '重新生成没有复用分页 canonical 正文加载器')
assert.match(regenerationSource, /getDataProcessPreview\(taskId, \{ page: 1, page_size: 500 \}\)[\s\S]*?for \(let page = 2; page <= pages;/, '重新生成没有分页加载全部现有切片')
assert.match(viewSource, /if \(hydrating\.value\) return/, '任务水合期间仍可能触发重置副作用')
assert.match(regenerationSource, /currentSignature !== originalPreviewConfigSignature\.value[\s\S]*?currentSignature === confirmedPreviewConfigSignature\.value/, '切分变更确认没有按原签名和已确认签名去重')
assert.match(regenerationSource, /if \(!bindings\.previewItems\.value\.length\) return true/, '没有现有切片时仍会弹出删除切片警告')
assert.ok(regenerationSource.includes('点击“开始生成”前,原生成结果和已发布数据会继续保留'), '切分变更警告没有说明开始生成前原结果与发布数据仍保留')
assert.doesNotMatch(regenerationSource, /原切片、原生成结果/, '切分变更警告不应承诺实际重切后仍保留原切片')
assert.doesNotMatch(regenerationSource, /修改预处理或切分配置将删除现有切片和生成结果/, '准备配置阶段仍误称会立即删除原数据')
assert.match(regenerationSource, /regenerateDataProcessTask\(sourceTaskId\.value,[\s\S]*?expected_updated_at: originalTaskUpdatedAt\.value/, '第二步没有携带最新乐观并发版本重新生成任务')
assert.match(regenerationSource, /originalTaskUpdatedAt\.value = regenerated\.task\.updated_at/, '重新生成成功后没有更新下一次提交的并发版本')
assert.match(regenerationSource, /if \(regenerationPrepared\.value\) \{[\s\S]*?getDataProcessTask\(sourceTaskId\.value\)[\s\S]*?originalTaskUpdatedAt\.value = latestTask\.updated_at/, '服务端中间操作后再次提交没有刷新任务并发版本')
assert.match(regenerationSource, /regenerationPrepared\.value = true/, '重新生成提交成功后没有记录服务端已变更状态')
assert.match(regenerationSource, /hydrateWorkspace\(regeneratedTask, !regenerated\.preview_invalidated\)/, '重新生成没有按 preview_invalidated 决定保留或清空切片')
assert.match(regenerationSource, /重新生成配置已保存,但工作区恢复失败/, '重新生成配置已保存但水合失败时缺少可恢复错误状态')
assert.match(sourceUploadWorkerSource, /return chunks\.join\(''\)/, '分页恢复源正文时不应额外插入换行')
assert.doesNotMatch(regenerationSource, /binaryDocument[\s\S]*?mapDataProcessSourceFile\(file, ''\)/, '二进制源正文加载失败时不能静默降级为空内容')
assert.match(nextFromModelSource, /if \(isRegeneration\.value\) \{[\s\S]*?prepareRegeneration\(taskPayload\(\)\)/, '重新生成每次从模型步骤继续时没有调用专用接口')
assert.doesNotMatch(nextFromModelSource, /isRegeneration\.value && !taskId\.value/, '重新生成提交一次后可能错误转为普通任务更新')
assert.doesNotMatch(regenerationSource, /leaveWarning|returnToDetail/, '返回列表已自动保存,不应保留旧版离开警告或返回详情分支')
assert.match(viewSource, /async function handleCancel\(\)[\s\S]*?returnToPreviousPage\(\)/, '第一步退出没有复用统一返回逻辑')
assert.match(viewSource, /async function returnToPreviousPage\(\)[\s\S]*?router\.push\(\{ name: 'data-process' \}\)/, '向导退出没有通过命名路由返回任务列表')
assert.match(regenerationSource, /async function confirmStartGeneration\([\s\S]*?开始重新生成?[\s\S]*?当前生成结果将被替换[\s\S]*?confirmText: '开始生成'[\s\S]*?syncPreviewChanges\(\)/, '开始生成前没有明确破坏性边界或同步预览修改')
assert.match(viewSource, /beforeGenerate: beforeStartGeneration[\s\S]*?async function beforeStartGeneration\(\)[\s\S]*?confirmStartGeneration\([\s\S]*?syncPreviewChanges/, '生成流程没有在真实开始前调用确认边界')
assert.match(generationSource, /const canStart = await bindings\.beforeGenerate\?\.\(\)[\s\S]*?if \(canStart === false\) return[\s\S]*?generateDataProcess\(taskId\)/, '用户取消开始生成时仍可能调用真实生成 API')
assert.match(viewSource, /v-if="initializationError"[\s\S]*?@click="loadRegenerationSource"[\s\S]*?重试加载原任务/, '原任务初始化失败后缺少安全重试入口')
assert.match(regenerationSource, /previewStatus = preservePreviews && count > 0 \? 'success' : 'waiting'/, '保留切片时没有跳过重切,或切片失效后未回到等待状态')
const previewOptionsStart = stateSource.indexOf('export function previewAffectingOptionsFor(')
const previewOptionsEnd = stateSource.indexOf('export function generationAffectingOptionsFor(', previewOptionsStart)
assert.ok(previewOptionsStart >= 0 && previewOptionsEnd > previewOptionsStart, '缺少预览影响配置签名函数')
const previewOptionsSource = stateSource.slice(previewOptionsStart, previewOptionsEnd)
for (const field of [
'preprocessOptions',
'chunkMethod',
'chunkSize',
'chunkOverlap',
'minChunkSize',
'semanticBreakpointPercentile',
'preserveTables',
'preserveCodeBlocks',
'preserveLists',
]) {
assert.ok(previewOptionsSource.includes(field), `预览签名缺少切分影响字段:${field}`)
}
for (const field of ['semanticEnrichment', 'qaPairsPerChunk', 'datasetSplit']) {
assert.ok(!previewOptionsSource.includes(field), `生成字段 ${field} 不应导致预览重建并丢失编辑`)
}
const generationOptionsStart = stateSource.indexOf('export function generationAffectingOptionsFor(')
const generationOptionsEnd = stateSource.length
assert.ok(generationOptionsStart >= 0 && generationOptionsEnd > generationOptionsStart, '缺少生成影响配置签名函数')
const generationOptionsSource = stateSource.slice(generationOptionsStart, generationOptionsEnd)
for (const field of [
'semanticEnrichment',
'qaPairsPerChunk',
'datasetSplit',
'generationModelId',
'generationPrompt',
'outputType',
'reasoningDetail',
'qualityFilterEnabled',
'filterLowQuality',
'filterShortContent',
'minOutputLength',
]) {
assert.ok(generationOptionsSource.includes(field), `生成签名缺少字段:${field}`)
}
assert.match(
viewSource,
/watch\(generationOptionsSignature,[\s\S]*?resetDownstream\(\)/,
'生成配置变化后没有仅失效下游结果',
)
for (const mutationFunction of [
'updatePreviewContent',
'restorePreviewItem',
'addPreviewItem',
'removePreviewItem',
]) {
const mutationStart = viewSource.indexOf(`function ${mutationFunction}`)
const mutationEnd = viewSource.indexOf('\nfunction ', mutationStart + 1)
const mutationSource = viewSource.slice(mutationStart, mutationEnd === -1 ? undefined : mutationEnd)
assert.ok(mutationSource.includes('resetDownstream()'), `预览变更 ${mutationFunction} 后没有失效旧生成结果`)
}
const updatePreviewContentStart = viewSource.indexOf('function updatePreviewContent(')
const updatePreviewContentEnd = viewSource.indexOf('\n}', updatePreviewContentStart)
const updatePreviewContentSource = viewSource.slice(updatePreviewContentStart, updatePreviewContentEnd + 2)
assert.match(updatePreviewContentSource, /isManualPreviewItem\(item\)/, '编辑预览内容仍未按稳定来源信息区分手动项')
assert.doesNotMatch(updatePreviewContentSource, /sourceStart == null/, '结构化来源记录编辑后仍会被误标为手动项')
const restorePreviewItemStart = viewSource.indexOf('function restorePreviewItem(')
const restorePreviewItemEnd = viewSource.indexOf('\n}', restorePreviewItemStart)
const restorePreviewItemSource = viewSource.slice(restorePreviewItemStart, restorePreviewItemEnd + 2)
assert.match(restorePreviewItemSource, /isManualPreviewItem\(item\)/, '恢复预览内容没有使用统一的手动项判定')
assert.doesNotMatch(restorePreviewItemSource, /sourceStart == null/, '结构化来源记录仍因缺少字符偏移而无法恢复')
assert.match(previewSource, /v-if="!isManualPreviewItem\(editingItem\)"/, '结构化来源记录的恢复原文按钮仍被错误隐藏')
assert.doesNotMatch(modelSource, /createResults\(/, '纯预览映射模块不应承担结果生成职责')
function findNextStyleBlockStart(source, startIndex) {
let quote = null
for (let index = startIndex; index < source.length; index += 1) {
const character = source[index]
const nextCharacter = source[index + 1]
if (quote) {
if (character === '\\') {
index += 1
} else if (character === quote) {
quote = null
}
continue
}
if (character === '/' && nextCharacter === '*') {
const commentEnd = source.indexOf('*/', index + 2)
index = commentEnd === -1 ? source.length : commentEnd + 1
continue
}
if (character === '/' && nextCharacter === '/') {
const commentEnd = source.indexOf('\n', index + 2)
index = commentEnd === -1 ? source.length : commentEnd
continue
}
if (character === '\'' || character === '"') {
quote = character
continue
}
if (character === '{') return index
}
return -1
}
function findStyleBlockEnd(source, blockStart) {
let depth = 0
let quote = null
for (let index = blockStart; index < source.length; index += 1) {
const character = source[index]
const nextCharacter = source[index + 1]
if (quote) {
if (character === '\\') {
index += 1
} else if (character === quote) {
quote = null
}
continue
}
if (character === '/' && nextCharacter === '*') {
const commentEnd = source.indexOf('*/', index + 2)
index = commentEnd === -1 ? source.length : commentEnd + 1
continue
}
if (character === '/' && nextCharacter === '/') {
const commentEnd = source.indexOf('\n', index + 2)
index = commentEnd === -1 ? source.length : commentEnd
continue
}
if (character === '\'' || character === '"') {
quote = character
continue
}
if (character === '{') {
depth += 1
} else if (character === '}' && --depth === 0) {
return index
}
}
return -1
}
function resolveNestedSelector(selector, parentSelector) {
if (!parentSelector) return selector
if (selector.includes('&')) return selector.replace(/&/g, parentSelector)
return `${parentSelector} ${selector}`
}
function collectStyleRules(source, parentSelector = '') {
const rules = []
let ruleStart = 0
let cursor = 0
while (cursor < source.length) {
const blockStart = findNextStyleBlockStart(source, cursor)
if (blockStart === -1) break
const blockEnd = findStyleBlockEnd(source, blockStart)
if (blockEnd === -1) break
const rawSelector = source.slice(ruleStart, blockStart).trim()
const declarations = source.slice(blockStart + 1, blockEnd)
if (rawSelector) {
const isAtRule = rawSelector.startsWith('@')
const selector = isAtRule ? rawSelector : resolveNestedSelector(rawSelector, parentSelector)
rules.push({ selector, declarations })
rules.push(...collectStyleRules(declarations, isAtRule ? parentSelector : selector))
}
cursor = blockEnd + 1
ruleStart = cursor
}
return rules
}
function directStyleDeclarations(source) {
let result = ''
let nestedDepth = 0
let quote = null
for (let index = 0; index < source.length; index += 1) {
const character = source[index]
const nextCharacter = source[index + 1]
if (quote) {
if (character === '\\') {
index += 1
} else if (character === quote) {
quote = null
}
if (nestedDepth === 0) result += ' '
continue
}
if (character === '/' && nextCharacter === '*') {
const commentEnd = source.indexOf('*/', index + 2)
index = commentEnd === -1 ? source.length : commentEnd + 1
if (nestedDepth === 0) result += ' '
continue
}
if (character === '/' && nextCharacter === '/') {
const commentEnd = source.indexOf('\n', index + 2)
index = commentEnd === -1 ? source.length : commentEnd
if (nestedDepth === 0) result += ' '
continue
}
if (character === '\'' || character === '"') {
quote = character
if (nestedDepth === 0) result += ' '
continue
}
if (character === '{') {
nestedDepth += 1
if (nestedDepth === 1) result += ' '
continue
}
if (character === '}') {
nestedDepth = Math.max(0, nestedDepth - 1)
if (nestedDepth === 0) result += ' '
continue
}
if (nestedDepth === 0) result += character
}
return result
}
const nestedUploadedFileItemsStyles = collectStyleRules(`
.upload-context {
.uploaded-file {
&-items {
max-height: 20rem;
}
@media (min-width: 1px) {
&-items {
overflow: auto;
}
}
@supports (display: grid) {
&-items {
overflow-x: hidden;
overflow-y: auto;
}
}
}
}
`).filter(({ selector }) => /(?:^|[^\w-])\.uploaded-file-items(?![\w-])/.test(selector))
assert.equal(
nestedUploadedFileItemsStyles.length,
3,
'嵌套的 &-items 选择器必须展开为 .uploaded-file-items且不能被 at-rule 上下文遮蔽',
)
for (const property of ['max-height', 'overflow', 'overflow-x', 'overflow-y']) {
assert.ok(
nestedUploadedFileItemsStyles.some(({ declarations }) =>
new RegExp(`(?:^|;)\\s*${property}\\s*:`, 'i').test(directStyleDeclarations(declarations)),
),
`嵌套 .uploaded-file-items 必须识别受限样式属性:${property}`,
)
}
for (const marker of [
'uploaded-file-list-header',
'uploaded-file-items',
'已选择 {{ uploadedFiles.length }} 个文件',
':title="file.name"',
]) {
assert.ok(sourceUploadSource.includes(marker), `源数据文件列表缺少:${marker}`)
}
assert.match(
sourceUploadSource,
/^const[ \t]+FILE_PAGE_SIZE[ \t]*=[ \t]*10[ \t]*;?[ \t]*$/m,
'文件分页大小必须固定为整数 10',
)
assert.match(
sourceUploadSource,
/const pagedUploadedFiles\s*=\s*computed\(\(\)\s*=>\s*\{\s*const start = \(currentFilePage\.value - 1\) \* FILE_PAGE_SIZE\s*return props\.uploadedFiles\.slice\(start, start \+ FILE_PAGE_SIZE\)\s*\}\)/,
'文件分页必须按当前页偏移切片完整文件列表',
)
assert.match(
sourceUploadSource,
/watch\(\(\)\s*=>\s*props\.uploadedFiles\.length,\s*\(newLength, oldLength\)\s*=>\s*\{[\s\S]*?if \(newLength > oldLength\)\s*\{\s*currentFilePage\.value = totalPages[\s\S]*?\}[\s\S]*?currentFilePage\.value = Math\.min\(currentFilePage\.value, totalPages\)[\s\S]*?\}\)/,
'文件数变化时必须新增跳至末页、删除回退到有效页',
)
const filePaginationTags = [...sourceUploadSource.matchAll(/<el-pagination\b[\s\S]*?\/>/g)]
assert.ok(filePaginationTags.length >= 1, '文件列表必须包含分页器')
for (const [filePaginationTag] of filePaginationTags) {
for (const attribute of [
'v-if="uploadedFiles.length > FILE_PAGE_SIZE"',
'v-model:current-page="currentFilePage"',
':page-size="FILE_PAGE_SIZE"',
':total="uploadedFiles.length"',
]) {
assert.ok(filePaginationTag.includes(attribute), `文件分页器缺少属性:${attribute}`)
}
}
const { descriptor: sourceUploadDescriptor } = parseSfc(sourceUploadSource, { filename: sourceUploadPath })
const uploadedFileItemsStyles = sourceUploadDescriptor.styles
.flatMap(({ content }) => collectStyleRules(content))
.filter(({ selector }) => /(?:^|[^\w-])\.uploaded-file-items(?![\w-])/.test(selector))
for (const { declarations } of uploadedFileItemsStyles) {
assert.doesNotMatch(
directStyleDeclarations(declarations),
/(?:^|;)\s*(?:max-height|overflow|overflow-x|overflow-y)\s*:/i,
'文件列表不能用内部滚动替代分页',
)
}
assert.match(
sourceUploadSource,
/<el-upload\s+v-if="uploadedFiles\.length === 0"[\s\S]*?<\/el-upload>\s*<section\s+v-else\s+class="uploaded-file-list"\s+aria-label="已上传文件列表">/,
'有文件状态缺少带 aria-label="已上传文件列表" 的语义列表容器',
)
assert.match(
sourceUploadSource,
/<div\s+class="uploaded-file-list-header">[\s\S]*?已选择 \{\{ uploadedFiles\.length \}\} 个文件[\s\S]*?正在逐个上传/,
'文件列表标题结构或文件数量文案缺失',
)
assert.match(
sourceUploadSource,
/<div\s+class="uploaded-file-items">\s*<div\s+v-for="file in pagedUploadedFiles"[^>]*class="uploaded-file">/,
'文件列表缺少分页后的文件行容器',
)
assert.match(
sourceUploadSource,
/<el-upload\s+v-if="uploadedFiles\.length === 0"\s+drag\s+multiple\s+:accept="uploadAccept"\s+:disabled="previewBuilding"\s+:auto-upload="false"\s+:show-file-list="false"\s+:on-change="\(file: UploadFile\) => emit\('file-change', file\)"[^>]*>/,
'无文件时未保留原有大拖拽上传区或上传配置',
)
assert.match(
sourceUploadSource,
/<el-upload\s+v-if="uploadedFiles\.length === 0"[\s\S]*?<template\s+#tip>[\s\S]*?processType === 'unstructured'/,
'无文件时大拖拽上传区缺少按处理类型展示的格式提示',
)
assert.ok(
sourceUploadSource.includes("? '.txt,.md,.markdown,.pdf,.docx,.pptx,.json,.jsonl,.ndjson'"),
'非结构化上传 accept 不完整',
)
assert.ok(
sourceUploadSource.includes(": '.json,.jsonl,.ndjson,.csv,.tsv,.xlsx'"),
'结构化上传 accept 不完整',
)
assert.ok(sourceUploadSource.includes('旧版 DOC/PPT 请先转换'), '非结构化格式提示没有说明旧版 DOC/PPT 需转换')
assert.ok(sourceUploadSource.includes('旧版 XLS 请先转换'), '结构化格式提示没有说明旧版 XLS 需转换')
assert.match(
sourceUploadSource,
/<div\s+class="uploaded-file-list-header">[\s\S]*?已选择 \{\{ uploadedFiles\.length \}\} 个文件[\s\S]*?<div\s+class="continue-upload">\s*<el-upload\s+multiple\s+:accept="uploadAccept"\s+:disabled="previewBuilding"\s+:auto-upload="false"\s+:show-file-list="false"\s+:on-change="\(file: UploadFile\) => emit\('file-change', file\)"[^>]*>\s*<el-button\s+size="small"\s+type="primary"\s+:disabled="previewBuilding">继续上传<\/el-button>\s*<\/el-upload>\s*<\/div>\s*<\/div>/,
'有文件时缺少标题右侧的继续上传触发器或上传配置',
)
assert.match(
sourceUploadSource,
/\.uploaded-file-list-header\s*\{[^}]*display:\s*flex[^}]*justify-content:\s*space-between/,
'文件列表标题未布局为右侧继续上传按钮',
)
assert.match(
sourceUploadSource,
/\.continue-upload\s+:deep\(\.el-upload\)\s*\{[^}]*width:\s*auto;?[^}]*margin-top:\s*0;?/,
'继续上传未覆盖内层上传节点的宽度和顶部间距',
)
assert.match(sourceUploadSource, /\.uploaded-file\s*\{[^}]*min-height:\s*48px/, '文件行没有保持 48px 最小高度')
assert.match(sourceUploadSource, /success:\s*\{ label: '切分完成', icon: 'fa-check-circle' \}/, '文件行缺少切分成功状态')
assert.match(sourceUploadSource, /previewBuilding/, '上传组件没有接收逐文件预览构建状态')
assert.match(sourceUploadSource, /sourceUploading/, '上传组件没有接收串行上传状态')
assert.match(
sourceUploadSource,
/<el-progress[\s\S]*?:percentage="getFileBarPercentage\(file\)"[\s\S]*?:indeterminate="isFileProcessing\(file\)"/,
'文件行缺少上传与切分阶段的独立进度',
)
for (const label of ['等待上传', '正在上传', '上传失败', '等待切分', '正在切分', '切分完成', '切分失败']) {
assert.ok(sourceUploadSource.includes(`label: '${label}'`), `文件行缺少状态:${label}`)
}
assert.match(sourceUploadSource, /file\.uploadProgress/, '文件行没有使用真实上传进度')
assert.match(sourceUploadSource, /file\.uploadError \|\| file\.previewError/, '文件行没有按阶段展示上传或切分失败原因')
assert.match(sourceUploadSource, /file\.previewError/, '文件行没有展示逐文件失败原因以支持重试')
assert.match(sourceUploadSource, /@click="emit\('remove-file', file\.uid\)"/, '文件行缺少 remove-file 删除动作')
const { descriptor } = parseSfc(viewSource, { filename: viewPath })
const template = descriptor.template?.content || ''
assert.equal((template.match(/class="wizard-primary-action"/g) || []).length, 1, '页面必须只有一个主操作入口')
assert.match(viewSource, /onBeforeUnmount\(\(\) => \{[\s\S]*?stopGenerationTimer\(\)[\s\S]*?\}\)/, '生成轮询计时器没有在卸载时清理')
assert.match(viewSource, /function scrollToStepTop/, '步骤切换后没有恢复页面顶部上下文')
assert.match(viewSource, /nextTick\(scrollToStepTop\)/, '步骤切换没有触发页面滚动复位')
assert.match(viewStyleSource, /\.wizard-content\s*\{[\s\S]*min-height:\s*400px/, '第一步内容区必须保留足够高度以显示底部操作栏')
assert.match(
layoutSource,
/&:has\(\.create-wizard-layout\)\s*\{[\s\S]*?overflow-y:\s*hidden[\s\S]*?\.page-canvas\s*\{[\s\S]*?flex:\s*1 1 auto[\s\S]*?min-height:\s*0/,
'创建任务页必须约束画布高度,避免底部操作栏被裁掉',
)
assert.match(previewSource, /height:\s*clamp\(560px,\s*calc\(100vh - 370px\),\s*720px\)/, '对照预览高度不足以展示切片正文')
console.log('数据处理六步向导回归检查通过')