hangshuo652/aurak
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
68 Commits 2 Branches 0 Tags
9fd503b42b74d90ead417bf47167a931916388c7
Commit Graph

68 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Developer 9fd503b42b feat: 考核系统升级 P0+P1+P2 — 体验/题库/配置增强 
P0 — 答题体验优化:
- 题序导航点：题目进度可视化，标记题目标记
- 标记回头检查: 点击🏷️按钮标记当前题，导航点变黄色
- 提交确认弹窗: 未答完时提交弹出确认对话框

P1 — 题库管理增强:
- QuestionBankItem 新增 tags 字段（多标签过滤）
- 新增 question_bank_templates 联表（题库跨模板复用）

P2 — 考试配置增强:
- AssessmentTemplate 新增字段:
  - attemptLimit (尝试次数限制)
  - scheduledStart/scheduledEnd (预约时段)
  - reviewMode (回顾模式: none/after_completion/per_question)
  - shuffleQuestions (每题随机排序)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-09 14:25:29 +08:00
Developer 5bbab82e68 test: 并发考核性能实验（20人同时考核） 
测试结果:
- Session ID 全部唯一: ✅ 20/20
- 异步出题完成: ✅ 20/20, 每题20题
- 维度分布正确: ✅ IDE:4/LLM:6/PROMPT:6/DEV_PATTERN:4
- 并发提交答案: ✅ 6人×4题全成功
- 题目重叠: ⚠️ 10.5%（题库仅~281题，20人需400槽位）
- 评分耗时: ⚠️ 15s不足以完成AI评分

结论: 并发场景下会话创建、出题、答题均正常，无数据竞争。
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-09 14:02:05 +08:00
Developer ebb8fbd298 docs: 统一 CLAUDE.md 和 README.md 文档体系 
CLAUDE.md（AI 参考）:
- 4 个 👉 AI 实现参考 标记，AI 可快速定位
- 完整的实体定义（TypeScript 代码块）
- 权限守卫流水线 + 解析链路
- 考核数据模型 + 出题算法伪代码
- API 认证流程 + 关键端点表
- 测试注意事项（React 受控输入/等待策略）
- 人可读用户指南（与 README 一致）

README.md（人可读）:
- 顶部引用 CLAUDE.md — AI 阅读指引
- 功能亮点表格式呈现
- 步骤式操作指南（用户/权限/模板/考试/租户）
- 轻量内容，详细技术参考指向 CLAUDE.md

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-09 13:29:28 +08:00
Developer 65ede9fcff docs: 全面更新文档体系 — AI指南 + 人可读说明书 
CLAUDE.md — AI 工作指南:
- 项目全景：目录结构/技术栈/端口
- 系统架构：前端路由/后端模块/认证流程
- 权限系统：三层角色/26项权限/守卫流水线/解析链路
- 考核系统：数据模型/出题算法/模板配置
- 测试脚本：7个Playwright测试说明
- 开发指南：启动/测试/重启/数据库管理
- 代码规范：TypeScript模式/权限装饰器/React约定
- Playwright测试技巧：React受控输入框/等待策略

README.md — 人可读英文说明书:
- 系统介绍 + 功能特性
- 完整使用指南（用户管理/权限管理/考核模板/组织考试）
- 核心流程说明（认证/出题/权限解析）
- 测试命令参考
- 项目结构 + 配置参考

README_ZH.md — 人可读中文说明书:
- 全面中文版本，包含所有新功能
- 步骤式操作指南，便于管理员使用

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-09 13:19:45 +08:00
Developer 1aee7e0baf fix: 出题分配算法重构 + 20题模板 + 非技术人员模板 
问题修复:
- Math.round 导致合计偏差（3题→4题，5题→6题等）
- 补充轮次破坏维度权重比例
- 各维度无题型比例控制

修正方案:
- floor + remainder 分配法，保证合计永远 = count
- 按weight降序分配余数，权重高的优先
- 不足时仅补充轮，但仍保持维度优先
- 补充轮日志记录各维度实际分配数

新增功能:
- 技术人员模板 20题 (PROMPT:30/LLM:30/IDE:20/DEV_PATTERN:20)
- 非技术人员模板 10题 (PROMPT:50/LLM:30/WORK_CAPABILITY:20)
  → 面向非技术角色，不考核IDE和开发范式
- 模板支持任意维度组合，可灵活配置

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-09 11:24:32 +08:00
Developer c166d298b8 fix(ui): 批量修复字号/间距/弹窗/表格 UI 问题（接续） 
- 密码修改弹窗: max-w-md→max-w-lg, py-3.5→py-3, text-[10px]→text-xs
- 角色选择按钮: flex-wrap + min-w-[100px] 防窄屏换行
- 用户表格: overflow-x-auto + min-w-[700px] 响应式
- 全部 42 处 text-[10px] → text-xs（标签/徽章/说明文字）
- PermissionSettingsView 剩余 2 处 text-[10px]→text-xs

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-09 10:39:48 +08:00
Developer ffe365201a fix(ui): 统一设计语言——颜色、字号、间距、控件样式 
UI 修正清单：
┌────────────────────────────────────────────────────────────┐
│ 问题                     │ 修复                          │
├──────────────────────────┼────────────────────────────────┤
│ 登录页 blue → 不一致     │ 统一改为 indigo 系             │
│ text-[10px] 标签难读     │ → text-xs (设置页所有标签)     │
│ text-[9px] 徽章太小      │ → text-xs (角色/租户/权限)    │
│ text-[14px] 输入框不统一  │ → text-sm (编辑弹窗)          │
│ py-3 vs py-3.5 不一致    │ → py-3 统一 (所有输入框)      │
│ 表格头 py-4 过大         │ → py-3                        │
│ 编辑弹窗 max-w-md 太窄   │ → max-w-lg (更宽松)           │
│ 操作列 opacity-0 隐藏     │ → opacity-60 始终可见         │
│ 原生 checkbox 不匹配      │ → 自定义圆角 checkbox + Check  │
│ 登录页 rounded-lg         │ → rounded-xl 统一             │
│ 登录按钮缺阴影            │ → 加 shadow-lg                │
└────────────────────────────────────────────────────────────┘

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-09 10:28:49 +08:00
Developer 7e741651db test: 系统性测试142项全通过 + 修复GET /users/:id缺失 
测试架构（10大类142项）:
┌──────────────────────────────────────────────────────┐
│ 1. 环境准备       4项  环境可达性 + 残留清理          │
│ 2. 身份认证      15项  登录/错误密码/空/篡改JWT      │
│ 3. 用户CRUD正常  11项  创建/查询/编辑/升降级/删除    │
│ 4. 用户CRUD异常  17项  重复/空/短密码/不存在/权限    │
│ 5. 边界测试       7项  并发/超长/空权限/幂等         │
│ 6. 权限矩阵RBAC  49项  3层角色权限/API校验/系统保护  │
│ 7. 租户隔离       1项  跨租户不可见                  │
│ 8. 缺陷回归       5项  系统角色保护/幂等删除          │
│ 9. 前端UI一致    22项  登录/导航/Tab/弹窗/3角色      │
│ 10.用户故事完整  14项  SA/TA/USER闭环/升降级即时生效 │
└──────────────────────────────────────────────────────┘

发现并修复:
- 系统角色权限可被任意修改（isSystem 保护缺失）
- GET /users/:id 端点不存在

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-09 10:01:04 +08:00
Developer a7e7c85ff6 fix: 系统角色权限保护 + 全角色全场景 E2E 测试（94项） 
缺陷修复：
- PermissionService.setRolePermissions 增加 isSystem 检查
  系统角色的权限不可被修改（之前可被任意改写）

测试覆盖（94项全部通过）：
- PHASE A: 身份认证（登录/错误密码/无效token/空凭据 8项）
- PHASE B: 三层角色权限边界（26/21/5 权限一致性 3项）
- PHASE C: 创建用户异常（重复/短密码/空字段/特殊字符 7项）
- PHASE D: 编辑&角色变更（改名/升降级/非法值/并发/跨角色 12项）
- PHASE E: 删除异常（删自己/admin/不存在/USER删/TA删 12项）
- PHASE F: 权限系统（角色CRUD/权限改/权限一致性/元数据 25项）
- PHASE G: 模块可达性（2项，非致命）
- PHASE H: 前端UI（admin/ta_admin/user1 三角色 22项）
- PHASE I: 边界缺陷（跨租户隔离/超长名 2项）

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-09 09:41:04 +08:00
Developer 64771f10ed test: 用户管理全生命周期测试（42项）覆盖异常case 
覆盖场景：
- 创建用户异常：重复用户名、密码太短、空字段
- 权限边界：USER 不能创建/查看/删除用户
- 角色变更：USER↔TENANT_ADMIN 切换后权限实时生效
- 删除异常：删自己、删 admin、删不存在用户
- UI 验证：角色列、编辑弹窗、权限管理页、权限矩阵

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-09 09:00:23 +08:00
Developer 9b4412792b feat: 用户管理页加角色列和角色编辑弹窗 
- 用户表新增「角色」列，显示超级管理员/管理员/用户徽章
- 编辑用户弹窗增加角色选择（USER/TENANT_ADMIN/SUPER_ADMIN）
- 角色选择时同步显示该角色的权限预览
- 保存时自动调用 PATCH /tenants/:id/members/:userId 更新角色
- 新增 test-permission-flow.mjs 三层用户权限测试脚本

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-08 23:37:13 +08:00
Developer ba33d517c1 feat: 分层 RBAC 权限管理系统 
后端:
- 新增 Role / RolePermission 实体（自动 seed 系统角色）
- PermissionService——通过 isAdmin / TenantMember 链路解析用户权限
- @Permission() 装饰器 + PermissionsGuard 守卫
- /api/permissions 和 /api/roles REST API
- UserController 内联 role 检查迁移到 @Permission()
- PermissionModule 全局注册

前端:
- usePermissions hook——获取当前用户权限集
- PermissionGate 组件级门控
- PermissionSettingsView——角色列表+权限矩阵编辑页面
- SettingsView 新增「权限管理」Tab（仅 admin 可见）
- 权限预览（26 项，7 分类）

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-08 23:25:22 +08:00
Developer c57c3028e2 fix: shuffleArray bug + Playwright多轮对话测试 + 初学者考核脚本 
- 修复 shuffleArray 返回新数组但调用处用 const 未接收返回值（3处）
- 新增 test-multiround.mjs Playwright 多轮对话测试（简答+追问全流程）
- 新增 do-assessment.mjs / check-result.mjs 考核体验脚本
- CLAUDE.md 增加 AI 工作流指令规则
- package.json 添加 playwright 依赖

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-08 22:34:04 +08:00
Developer 0b2c6563ba Add IDE题库 50 questions + PROMPT/LLM/DEV_PATTERN 34 questions 2026-06-03 21:12:11 +08:00
Developer 6d9acd7252 fix: MC options display, question selection, timeout handling, and grading prompts 2026-06-03 20:58:19 +08:00
Developer a71bde3452 add IDE question bank seed script (50 questions for L1 IDE协作开发) 
- Copilot: 12 questions (intelligent completion, Chat modes, CLI)
- Claude Code: 12 questions (interaction methods, model selection, CLI commands)
- OpenCode: 19 questions (overview, installation, Plan/Build, commands, models)
- Debug: 7 questions (/fix, /debug, /tests, /explain commands)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-03 20:55:38 +08:00
Developer 6e569ff478 fix: skip content check when bank questions available, early generator return 2026-05-23 22:32:51 +08:00
Developer a83de861dd fix: replace PDF with HTML report (fontkit unavailable) 2026-05-21 16:30:36 +08:00
Developer 0b0da09d4b fix: use pdf-lib embedFont with proper pagination for CJK PDF 2026-05-21 16:12:38 +08:00
Developer d7cd5641d7 fix: rewrite PDF generator using pdf-lib native embedFont for CJK support 2026-05-21 15:59:01 +08:00
Developer c53f26a07e perf: trim grader prompts ~40% to reduce LLM latency 2026-05-21 15:52:33 +08:00
Developer b15e821252 feat: enriched certificate with template name, dimension scores, question details + Modal UI 
- generateCertificate: return templateName, questionDetails, dimensionScores
- Frontend: replace alert() with certificate Modal showing level, scores, dimensions, questions
- Status label: change from '已验证' to '合格'
 2026-05-21 15:42:59 +08:00
Developer 990b8c7b83 fix: forward passed flag in SSE final events 2026-05-21 15:24:36 +08:00
Developer f8df92c36b fix: forward finalScore in submitAnswerStream final event 2026-05-21 15:19:10 +08:00
Developer 51f2a41cc3 fix: determineLevel uses 0-10 scale thresholds instead of 0-100 2026-05-21 15:07:23 +08:00
Developer 0a3a8a2e32 fix: send accumulated answers to LLM grader for follow-up context 
- Grader now passes all rounds of user answers to LLM (tagged 第N轮回答)
- LLM can see what was already answered and avoid redundant follow-ups
- Updated all three language prompts with multi-round guidance
 2026-05-21 14:41:57 +08:00
Developer 9303d7ac64 fix: auto-submit answer on timeout instead of blocking 
- Timeout triggers forced submission of current answer (or empty)
- Prevents assessment from hanging when time expires
- autoSubmitted flag prevents duplicate submissions
 2026-05-21 14:32:01 +08:00
Developer 02f4ab23f7 feat: LLM-generated adaptive follow-up questions 
- Grader: LLM outputs follow_up_question targeting uncovered keyPoints
- Remove static followupHints usage in grading flow
- maxFollowUps sourced from question.maxFollowUps (hints.length)
- Clean answerKey: remove followupHints field
- Three-language prompt update with examples and bad examples
- Grader spec: add follow_up_question to mock responses
 2026-05-21 14:18:14 +08:00
Developer 7fd2a4cda2 fix: option display + partial credit grading 
- Option display: use slice(1) instead of regex to strip letter prefix
- Grader prompts: add explicit partial credit guidance (5-7 for partial, 0-2 only for off-target)
 2026-05-21 13:13:21 +08:00
Developer 7b1103903f fix: remove prefix from followup hint, use raw text 2026-05-21 13:00:03 +08:00
Developer 3cc3b28471 fix: broader regex to strip conditional prefix from followup hints 2026-05-21 12:57:04 +08:00
Developer 5c82c75a09 fix: strip option letter prefix in QuestionBankDetailView 
Consistent with AssessmentView, now strips A./B./C./D. prefix
from option text before displaying alongside letter badge.
 2026-05-21 12:48:35 +08:00
Developer 24ffc028e2 fix: shuffle choice options per session + clean followup hints 
- Options shuffled with correctAnswer remapped at session creation
- Followup hints strip conditional prefix (如果只说了XX，追问：)
 2026-05-21 12:42:52 +08:00
Developer 734c0129d8 chore: remove test artifact 2026-05-21 11:53:47 +08:00
Developer 1224a74e63 fix: natural follow-up conversation flow 
- Grader: separate followup hint from scoring feedback
- Interviewer: use followup hint directly without prefix/suffix
- Restored standard and choice question presentation paths
 2026-05-21 11:53:24 +08:00
Developer c015ea3697 fix: shuffle bank questions + grader LLM error resilience 
- selectQuestions: shuffle final result for random question order
- grader: wrap LLM invoke in try-catch, default score 5 on failure
- grader: inner try-catch for JSON parse errors, graceful fallback
 2026-05-21 11:33:17 +08:00
Developer 240aea24aa fix: linkedGroupIds null check in validateRequiredFields 
null !== undefined was true, causing false validation failure on templates
without linked groups. Changed to != null check.
 2026-05-21 11:17:45 +08:00
Developer 54762ca299 fix: passingScore scaling and dimensions propagation 
- Frontend: divide by 10 on load, multiply by 10 on send (UI:0-10, DB:0-100)
- Backend: include template dimensions in session templateData snapshot
 2026-05-21 11:07:07 +08:00
Developer eba30517a6 fix: remove bank PUBLISHED guard from selectQuestions 
selectQuestions now only checks item-level PUBLISHED status.
startSession already handles bank detection by counting published items.
This fixes assessment always falling back to LLM generation.
 2026-05-21 10:26:19 +08:00
Developer 35b1c6c37d feat: judgment-anchored grading and per-question results 
- Grader: inject judgment as pass criteria anchor in LLM prompt
- Grader: use followupHints for follow-up direction (not generic text)
- Grader: follow-up limit from followupHints.length instead of hardcoded 2
- Session: correctAnswer/judgment stored in questions, stripped during assessment
- Frontend: per-question results panel with choice ✅/❌ + judgment display
 2026-05-21 10:18:15 +08:00
Developer 3993099907 feat: end-to-end choice question support in assessment pipeline 
- Data pathway: flow options through questions, answerKey in graph state
- Interviewer: format MULTIPLE_CHOICE with A/B/C/D options
- Grader: instant choice scoring (zero LLM), compare correctAnswer
- AssessmentView: render choice buttons vs textarea based on questionType
- Security: sanitizeStateForClient strips correctAnswer/judgment/answerKey
- Bank detection: check PUBLISHED items (not PUBLISHED bank status)
- Batch UI: select all / batch approve / batch reject on detail view
 2026-05-21 10:06:33 +08:00
Developer 57898f939c fix: add status guards to prevent data loss 
- create: auto-delete REJECTED→throw error; add tenantId filter
- remove: forbid PUBLISHED bank deletion
- removeItem: forbid PUBLISHED item deletion
- generateQuestions: restrict to DRAFT status only
- frontend: render MULTIPLE_CHOICE options/judgment/followupHints
- frontend: add judgment and followupHints to QuestionBankItem type
- add 12 service guard tests (109 total)
 2026-05-21 08:55:35 +08:00
Developer e782d180d7 feat: support choice+open dual question generation with judgment anchors 
- Add judgment and followupHints fields to QuestionBankItem entity
- Rewrite generateQuestions prompt for 3:7 choice:open ratio
- Extract parseGeneratedQuestion function with type-aware parsing
- Add 29 unit tests: 14 prompt content + 15 parse logic
- Total: 97 tests passing (59 baseline + 38 new)
 2026-05-21 01:04:08 +08:00
Developer 17ddfa83bf Question generation: scenario-based 3-step prompt with technique labeling, key_points constrained to KB source, temperature 0.1. Generator node: two-step extraction prompt for assessment flow. 2026-05-20 17:33:28 +08:00
Developer 83483d8117 F1-F10: audit fixes (dimension normalize, passingScore scale, DB defaults, onDelete, item status filter, timeout event type, userId privacy) + generator.node.ts strict prompt rules (anti-hallucination) 2026-05-20 11:13:37 +08:00
Developer 29bac74b58 M3: console.log -> Logger + UI redesign (QuestionBank) + S7/A9/A10/A11/U11 bug fixes + #1/#2/#3/#4 enhancements + i18n for QuestionBank pages 2026-05-19 16:57:45 +08:00
Developer 5b5f14674d fix: minor issues from code review 
(M1) DTO: @IsObject({ each: true }) on dimensions array
(M2) audit log: add missing tenantId in submitAnswer
(M3) console.log -> this.logger in controller + service
 2026-05-19 10:22:18 +08:00
Developer 82a9e75842 fix: code review — 7 issues resolved 
(C1) Add dimensionScores/radarData/passed columns to AssessmentSession
(C2) Mock DataSource in service.spec.ts + app.e2e-spec.ts
(C3) Mock AuditLogService in controller.spec.ts
(C4) Rewrite deleteSession tests for dataSource.transaction
(I1) batchDeleteSessions uses transaction with certificate cleanup
(I2) extractDimensionScores reads from session property
(I3/I5) PDF generator supports multi-page + newline splitting
(I4) findOne inside transaction uses deleteCondition
 2026-05-19 10:06:30 +08:00
Developer 7f8e7214b3 P3-02-03-04: audit log, batch ops, transactions 
P3-02: audit-log.entity + service, manual logging in controller
  (startSession, submitAnswer, deleteSession, review, forceEnd)
P3-03: POST batch-delete, POST batch-export endpoints + service methods
P3-04: DataSource.transaction for deleteSession + reviewAssessment,
  graph state cleanup on session delete
 2026-05-19 09:52:31 +08:00
Developer eb0798de5b P2-1: remove dead cost-control module (3 files) 
P2-2: switch TypeORM to autoLoadEntities: true

Remove unused vision-pipeline-cost-aware.service.ts,
cost-control.service.ts and its orphan module.
Switch explicit entities[] list to autoLoadEntities.
 2026-05-19 09:39:41 +08:00