feat(rvw): implement Skills architecture (Day 7-10)
- Add Skills core framework (types, registry, executor, profile, context) - Implement DataForensicsSkill with DI, path security, graceful degradation - Implement EditorialSkill and MethodologySkill wrapping existing services - Extend ExtractionClient with IExtractionClient interface and analyzeDocx - Refactor reviewWorker to support V1/V2 architecture switching - Add Zod config validation and generic type support - Update development docs and module status Day 7: Skills core framework (~700 lines) Day 8: DataForensicsSkill + ExtractionClient extension (~400 lines) Day 9: EditorialSkill + MethodologySkill (~350 lines) Day 10: ReviewWorker integration (~280 lines) Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
@@ -0,0 +1,331 @@
|
||||
# RVW V2.0 Day 7-10 开发记录
|
||||
|
||||
**日期**: 2026-02-18
|
||||
**开发阶段**: Week 2 - Day 7-10
|
||||
**开发主题**: Skills 架构核心框架 + Skill 实现
|
||||
**开发状态**: ✅ 完成
|
||||
|
||||
---
|
||||
|
||||
## 1. 开发背景
|
||||
|
||||
### 1.1 任务目标
|
||||
|
||||
根据 RVW V2.0 Skills 架构技术设计文档,Day 7-10 的主要任务是实现完整的 Skills 框架:
|
||||
|
||||
- **Day 7**: Skills 核心框架(types, registry, executor, profile, context)
|
||||
- **Day 8**: DataForensicsSkill(依赖注入、路径安全)
|
||||
- **Day 9**: EditorialSkill + MethodologySkill
|
||||
- **Day 10**: ReviewWorker 改造(集成 SkillExecutor)
|
||||
|
||||
### 1.2 审查意见整合
|
||||
|
||||
Day 7-10 开发前,团队提交了《RVW V2.0 Skills 架构深度审查报告》,核心建议已整合:
|
||||
|
||||
| 建议 | 采纳情况 | 实现位置 |
|
||||
|------|----------|----------|
|
||||
| 使用 Zod 进行配置验证 | ✅ 采纳 | `types.ts`, `BaseSkill.ts` |
|
||||
| 依赖注入(ExtractionClient) | ✅ 采纳 | `DataForensicsSkill.ts` |
|
||||
| 路径安全检查 | ✅ 采纳 | `DataForensicsSkill.ts` |
|
||||
| onSkillComplete 回调预留 | ✅ 采纳 | `executor.ts` |
|
||||
| 泛型设计 | ✅ 采纳 | `types.ts`, `BaseSkill.ts` |
|
||||
| 增量持久化 | ⏸️ 保留 V2.1 | 接口已预留 |
|
||||
|
||||
---
|
||||
|
||||
## 2. 开发成果
|
||||
|
||||
### 2.1 新增文件清单
|
||||
|
||||
| 文件 | 功能 | 代码行数 |
|
||||
|------|------|---------|
|
||||
| `skills/core/types.ts` | 类型定义 + Zod Schema | ~280 |
|
||||
| `skills/core/registry.ts` | Skill 注册表(单例) | ~130 |
|
||||
| `skills/core/executor.ts` | Pipeline 执行引擎 | ~270 |
|
||||
| `skills/core/profile.ts` | Profile 解析器 + 预设配置 | ~200 |
|
||||
| `skills/core/context.ts` | 上下文构建器 | ~130 |
|
||||
| `skills/core/index.ts` | Core 模块导出 | ~25 |
|
||||
| `skills/library/BaseSkill.ts` | Skill 基类(Zod 验证) | ~120 |
|
||||
| `skills/library/DataForensicsSkill.ts` | 数据侦探 Skill | ~200 |
|
||||
| `skills/library/EditorialSkill.ts` | 稿约规范性 Skill | ~150 |
|
||||
| `skills/library/MethodologySkill.ts` | 方法学评估 Skill | ~170 |
|
||||
| `skills/library/index.ts` | Library 模块导出 | ~50 |
|
||||
| `skills/index.ts` | 模块主入口 | ~10 |
|
||||
|
||||
**总计**: 12 个新文件,约 1735 行代码
|
||||
|
||||
### 2.2 修改文件清单
|
||||
|
||||
| 文件 | 修改内容 |
|
||||
|------|---------|
|
||||
| `common/document/ExtractionClient.ts` | 新增 `IExtractionClient` 接口 + `analyzeDocx` 方法 |
|
||||
| `modules/rvw/workers/reviewWorker.ts` | 改造为 Skills 架构,支持 V1/V2 切换 |
|
||||
|
||||
### 2.3 目录结构
|
||||
|
||||
```
|
||||
backend/src/modules/rvw/skills/
|
||||
├── core/ # 核心框架(未来移至 common/skills)
|
||||
│ ├── types.ts # 类型定义 + Zod Schema
|
||||
│ ├── registry.ts # Skill 注册表
|
||||
│ ├── executor.ts # Pipeline 执行引擎
|
||||
│ ├── profile.ts # Profile 配置解析
|
||||
│ ├── context.ts # 上下文管理
|
||||
│ └── index.ts # 统一导出
|
||||
├── library/ # Skill 实现库
|
||||
│ ├── BaseSkill.ts # Skill 基类
|
||||
│ ├── DataForensicsSkill.ts # 数据侦探
|
||||
│ ├── EditorialSkill.ts # 稿约规范性
|
||||
│ ├── MethodologySkill.ts # 方法学评估
|
||||
│ └── index.ts # 注册入口
|
||||
└── index.ts # 模块主入口
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 3. 技术实现详情
|
||||
|
||||
### 3.1 核心类型定义 (types.ts)
|
||||
|
||||
#### 3.1.1 泛型设计
|
||||
|
||||
```typescript
|
||||
// 基础上下文(通用)
|
||||
interface BaseSkillContext<TProfile = unknown> {
|
||||
taskId: string;
|
||||
userId?: string;
|
||||
previousResults: SkillResult[];
|
||||
profile: TProfile;
|
||||
}
|
||||
|
||||
// RVW 扩展字段
|
||||
interface RvwContextExtras {
|
||||
documentPath: string;
|
||||
documentContent: string;
|
||||
tables?: TableData[];
|
||||
methods?: string[];
|
||||
forensicsResult?: ForensicsResult;
|
||||
}
|
||||
|
||||
// RVW 完整上下文
|
||||
interface SkillContext extends BaseSkillContext<JournalProfile>, RvwContextExtras {}
|
||||
```
|
||||
|
||||
#### 3.1.2 Zod 配置 Schema
|
||||
|
||||
```typescript
|
||||
const DataForensicsConfigSchema = z.object({
|
||||
checkLevel: z.enum(['L1', 'L1_L2', 'L1_L2_L25']).default('L1_L2_L25'),
|
||||
tolerancePercent: z.number().min(0).max(1).default(0.1),
|
||||
});
|
||||
|
||||
const EditorialConfigSchema = z.object({
|
||||
standard: z.enum(['default', 'chinese-core', 'international']).default('default'),
|
||||
maxContentLength: z.number().default(100000),
|
||||
});
|
||||
```
|
||||
|
||||
### 3.2 执行引擎 (executor.ts)
|
||||
|
||||
#### 3.2.1 核心功能
|
||||
|
||||
- **Pipeline 执行**:按 Profile 配置顺序执行 Skills
|
||||
- **超时熔断**:可配置超时时间,默认 30 秒
|
||||
- **故障隔离**:单个 Skill 失败不影响整体(可配置)
|
||||
- **上下文传递**:前置 Skill 结果自动注入后续上下文
|
||||
- **回调扩展点**:`onSkillComplete` 预留增量持久化
|
||||
|
||||
#### 3.2.2 执行流程
|
||||
|
||||
```
|
||||
Profile.pipeline.forEach(item => {
|
||||
1. 检查 enabled 状态
|
||||
2. 获取 Skill 实例
|
||||
3. 执行 canRun 前置检查
|
||||
4. 带超时执行 skill.run()
|
||||
5. 调用 onSkillComplete 回调(V2.1)
|
||||
6. 更新上下文 previousResults
|
||||
7. 检查 continueOnError 策略
|
||||
})
|
||||
```
|
||||
|
||||
### 3.3 Profile 配置 (profile.ts)
|
||||
|
||||
#### 3.3.1 预设 Profiles
|
||||
|
||||
| Profile ID | 名称 | Pipeline | 特点 |
|
||||
|------------|------|----------|------|
|
||||
| `default` | 通用期刊配置 | Forensics → Editorial → Methodology | 标准模式 |
|
||||
| `chinese-core` | 中文核心期刊 | 同上 | 严格模式,失败即停止 |
|
||||
| `quick-forensics` | 快速数据侦探 | 仅 Forensics | 快速预览 |
|
||||
|
||||
#### 3.3.2 V1 兼容
|
||||
|
||||
```typescript
|
||||
// 将 V1.0 的 selectedAgents 映射到 V2.0 Skills
|
||||
const AGENT_TO_SKILL_MAP = {
|
||||
'editorial': 'EditorialSkill',
|
||||
'methodology': 'MethodologySkill',
|
||||
'forensics': 'DataForensicsSkill',
|
||||
};
|
||||
|
||||
// ProfileResolver.resolveFromAgents(['editorial', 'methodology'])
|
||||
// → 动态生成包含这些 Skills 的 Profile
|
||||
```
|
||||
|
||||
### 3.4 DataForensicsSkill 安全设计
|
||||
|
||||
#### 3.4.1 路径白名单
|
||||
|
||||
```typescript
|
||||
const ALLOWED_PATH_PREFIXES = [
|
||||
'/app/uploads/', // Docker 容器
|
||||
'D:\\MyCursor\\', // 开发环境
|
||||
'/tmp/rvw-uploads/', // 临时目录
|
||||
];
|
||||
|
||||
canRun(context: SkillContext): boolean {
|
||||
// 安全检查:路径白名单
|
||||
const isPathAllowed = ALLOWED_PATH_PREFIXES.some(prefix =>
|
||||
context.documentPath.startsWith(prefix)
|
||||
);
|
||||
|
||||
// 检查路径遍历
|
||||
if (context.documentPath.includes('..')) {
|
||||
return false; // 拒绝
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
#### 3.4.2 依赖注入
|
||||
|
||||
```typescript
|
||||
class DataForensicsSkill {
|
||||
private readonly extractionClient: IExtractionClient;
|
||||
|
||||
constructor(client?: IExtractionClient) {
|
||||
this.extractionClient = client || extractionClient;
|
||||
}
|
||||
}
|
||||
|
||||
// 测试时可注入 Mock
|
||||
const mockClient: IExtractionClient = { ... };
|
||||
const skill = new DataForensicsSkill(mockClient);
|
||||
```
|
||||
|
||||
### 3.5 ReviewWorker 改造
|
||||
|
||||
#### 3.5.1 架构切换
|
||||
|
||||
```typescript
|
||||
// 环境变量控制
|
||||
const USE_SKILLS_ARCHITECTURE = process.env.RVW_USE_SKILLS !== 'false';
|
||||
|
||||
// 运行时自动选择
|
||||
if (USE_SKILLS_ARCHITECTURE) {
|
||||
// V2.0 Skills 架构
|
||||
const profile = ProfileResolver.resolveFromAgents(agents);
|
||||
const summary = await executor.execute(profile, context);
|
||||
} else {
|
||||
// V1.0 Legacy 架构
|
||||
editorialResult = await reviewEditorialStandards(...);
|
||||
methodologyResult = await reviewMethodology(...);
|
||||
}
|
||||
```
|
||||
|
||||
#### 3.5.2 结果存储
|
||||
|
||||
```typescript
|
||||
// Skills 执行摘要存储到 picoExtract 字段(暂时复用)
|
||||
const skillsContext = {
|
||||
version: '2.0',
|
||||
executedAt: new Date().toISOString(),
|
||||
summary: {
|
||||
overallStatus: skillsSummary.overallStatus,
|
||||
totalSkills: skillsSummary.totalSkills,
|
||||
successCount: skillsSummary.successCount,
|
||||
errorCount: skillsSummary.errorCount,
|
||||
},
|
||||
forensicsResult: skillsSummary.results.find(r => r.skillId === 'DataForensicsSkill')?.data,
|
||||
};
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 4. 已知问题
|
||||
|
||||
### 4.1 数据库迁移阻塞
|
||||
|
||||
**问题**:尝试添加 `contextData` 字段时,Prisma migrate 报错(历史迁移问题)
|
||||
|
||||
**临时方案**:将 Skills 执行摘要存储到现有的 `picoExtract` JSON 字段
|
||||
|
||||
**后续计划**:修复历史迁移后,添加专用 `context_data` 字段
|
||||
|
||||
### 4.2 Python Forensics API
|
||||
|
||||
**状态**:`analyzeDocx` 方法已添加到 `ExtractionClient`,但 Python 端 API (`/api/v1/forensics/analyze`) 尚未实现
|
||||
|
||||
**后续计划**:Week 3 实现 Python 端完整 API
|
||||
|
||||
---
|
||||
|
||||
## 5. 测试状态
|
||||
|
||||
### 5.1 TypeScript 编译
|
||||
|
||||
✅ 无 Lint 错误
|
||||
|
||||
### 5.2 集成测试
|
||||
|
||||
⏳ 待 Python API 完成后进行端到端测试
|
||||
|
||||
---
|
||||
|
||||
## 6. 后续计划
|
||||
|
||||
### 6.1 Week 3 计划
|
||||
|
||||
| 任务 | 优先级 | 说明 |
|
||||
|------|--------|------|
|
||||
| Python Forensics API | P0 | 实现 `/api/v1/forensics/analyze` |
|
||||
| 前端表格渲染 | P1 | TaskDetail 页面展示提取的表格 |
|
||||
| 问题高亮 | P1 | 根据 R1C1 坐标高亮问题单元格 |
|
||||
| 端到端测试 | P1 | 完整流程测试 |
|
||||
|
||||
### 6.2 V2.1 规划
|
||||
|
||||
| 功能 | 说明 |
|
||||
|------|------|
|
||||
| `contextData` 专用字段 | 修复迁移后添加 |
|
||||
| 增量持久化 | 实现 `onSkillComplete` 回调 |
|
||||
| Profile 数据库存储 | 支持用户自定义 Profile |
|
||||
| 更多 Skills | 如引用格式检查、图表检查等 |
|
||||
|
||||
---
|
||||
|
||||
## 7. 变更日志
|
||||
|
||||
| 时间 | 变更内容 |
|
||||
|------|---------|
|
||||
| 2026-02-18 09:00 | 开始 Day 7 开发 |
|
||||
| 2026-02-18 09:30 | 创建 skills 目录结构 |
|
||||
| 2026-02-18 10:00 | 完成 types.ts(含 Zod Schema) |
|
||||
| 2026-02-18 10:30 | 完成 registry.ts |
|
||||
| 2026-02-18 11:00 | 完成 executor.ts(含超时熔断) |
|
||||
| 2026-02-18 11:30 | 完成 profile.ts(含 3 个预设 Profile) |
|
||||
| 2026-02-18 12:00 | 完成 context.ts |
|
||||
| 2026-02-18 13:00 | 完成 BaseSkill.ts |
|
||||
| 2026-02-18 13:30 | 扩展 ExtractionClient(IExtractionClient + analyzeDocx) |
|
||||
| 2026-02-18 14:00 | 完成 DataForensicsSkill(含路径安全) |
|
||||
| 2026-02-18 14:30 | 完成 EditorialSkill |
|
||||
| 2026-02-18 15:00 | 完成 MethodologySkill |
|
||||
| 2026-02-18 15:30 | 完成 Skills 注册入口 |
|
||||
| 2026-02-18 16:00 | 改造 reviewWorker |
|
||||
| 2026-02-18 16:30 | 修复 Lint 错误,处理数据库迁移问题 |
|
||||
| 2026-02-18 17:00 | 更新开发文档 |
|
||||
|
||||
---
|
||||
|
||||
*开发记录生成时间: 2026-02-18*
|
||||
*RVW V2.0 Skills 架构*
|
||||
Reference in New Issue
Block a user