docs(docs): 新增本地启动/多AI协作/CI-CD 使用手册
- local-dev-runbook.md: 8 章本地启动手册(环境依赖/端口分配/开发模式/生产模式/裸机运行/常见问题) - multi-ai-collaboration.md: 15 章多AI协作文档(角色定义/模块分工/分支策略/PR流程/审核合并/跨模块变更) - cicd-runbook.md: 10 章 CI/CD 使用手册(架构总览/一次性配置/日常使用/镜像管理/部署验证/回滚) - README.md: 文档清单新增三个 runbook 链接
This commit is contained in:
@@ -38,6 +38,9 @@ pnpm dev
|
||||
- [编码规范](docs/standards/coding-standards.md)
|
||||
- [UI 设计系统](docs/standards/ui-design-system.md)
|
||||
- [Git 工作流](docs/standards/git-workflow.md)
|
||||
- [本地启动手册](docs/standards/local-dev-runbook.md)
|
||||
- [多 AI 协作指南](docs/standards/multi-ai-collaboration.md)
|
||||
- [CI/CD 使用手册](docs/standards/cicd-runbook.md)
|
||||
- [已知问题](docs/troubleshooting/known-issues.md)
|
||||
- [项目规则](.trae/rules/project_rules.md)
|
||||
- [迁移指南](MIGRATION_GUIDE.md)
|
||||
|
||||
445
docs/standards/cicd-runbook.md
Normal file
445
docs/standards/cicd-runbook.md
Normal file
@@ -0,0 +1,445 @@
|
||||
# CI/CD 使用手册(CI/CD Runbook)
|
||||
|
||||
> 版本:1.0
|
||||
> 日期:2026-07-08
|
||||
> 适用范围:Edu 微服务项目(Gitea Actions + Gitea Container Registry + Runner 直接部署)
|
||||
> 关联文档:[project_rules §15](../../.trae/rules/project_rules.md)、[本地启动手册](./local-dev-runbook.md)、[多 AI 协作指南](./multi-ai-collaboration.md)
|
||||
|
||||
---
|
||||
|
||||
## 1. 架构总览
|
||||
|
||||
```
|
||||
开发 AI 推送 PR 协调 AI 合并到 main
|
||||
│ │
|
||||
▼ ▼
|
||||
┌─────────────┐ ┌─────────────┐
|
||||
│ ci-*.yml │ │ docker.yml │
|
||||
│ lint+test+ │ │ build+push │
|
||||
│ build+scan │ │ 镜像到 Gitea│
|
||||
└─────────────┘ │ Registry │
|
||||
└──────┬──────┘
|
||||
│
|
||||
▼
|
||||
┌─────────────┐
|
||||
│ deploy.yml │
|
||||
│ Runner 直接 │
|
||||
│ docker │
|
||||
│ compose up │
|
||||
└─────────────┘
|
||||
```
|
||||
|
||||
### 1.1 组件清单
|
||||
|
||||
| 组件 | 说明 |
|
||||
| ---------------------------- | --------------------------------------------------------------- |
|
||||
| **Gitea** | `git.eazygame.cn`,代码托管 + Actions + Container Registry |
|
||||
| **Gitea Actions** | CI 运行器,兼容 GitHub Actions 语法 |
|
||||
| **Gitea Container Registry** | 镜像仓库,地址 `git.eazygame.cn/xiner/edu/<service>:<tag>` |
|
||||
| **Deploy Runner** | 跑在服务器上的 self-hosted runner,标签 `[self-hosted, deploy]` |
|
||||
| **服务器 MySQL** | 已有容器,端口 3306,容器名 `edu-mysql`(假设) |
|
||||
| **服务器 Redis** | 已有容器,端口 6379,容器名 `edu-redis`(假设) |
|
||||
|
||||
### 1.2 流水线文件
|
||||
|
||||
| 文件 | 触发 | 作用 |
|
||||
| -------------------------------- | ------------------------ | ---------------------------------------------------------- |
|
||||
| `.github/workflows/ci-ts.yml` | TS 路径变更 | lint + typecheck + test + build + arch:scan + docker-build |
|
||||
| `.github/workflows/ci-go.yml` | Go 路径变更 | vet + build + test + docker-build |
|
||||
| `.github/workflows/ci-py.yml` | Python 路径变更 | ruff + pytest(P4 阶段) |
|
||||
| `.github/workflows/ci-proto.yml` | proto 变更 | buf lint + buf breaking |
|
||||
| `.github/workflows/docker.yml` | push main / tag `v*` | 构建并推送镜像到 Gitea Registry |
|
||||
| `.github/workflows/deploy.yml` | docker.yml 完成后 / 手动 | Runner 执行 docker compose up |
|
||||
|
||||
---
|
||||
|
||||
## 2. 一次性配置(SRE AI 或人类执行)
|
||||
|
||||
### 2.1 启用 Gitea Actions
|
||||
|
||||
在 Gitea 仓库 `xiner/Edu` 设置:
|
||||
|
||||
1. **仓库设置 → Actions → 启用**
|
||||
2. 安装 actrunner(若未安装):
|
||||
```bash
|
||||
# 在服务器上
|
||||
wget https://gitea.com/gitea/act_runner/raw/branch/main/act_runner
|
||||
chmod +x act_runner
|
||||
./act_runner register --instance https://git.eazygame.cn --token <TOKEN>
|
||||
```
|
||||
3. 配置 runner 标签为 `self-hosted,deploy`(部署用)和 `self-hosted,linux`(CI 用)
|
||||
|
||||
### 2.2 启用 Gitea Container Registry
|
||||
|
||||
Gitea 1.20+ 自带 Container Registry,无需额外启用。确认:
|
||||
|
||||
```bash
|
||||
# 测试登录
|
||||
docker login git.eazygame.cn -u <你的用户名>
|
||||
# 输入密码或 token
|
||||
```
|
||||
|
||||
### 2.3 准备服务器网络
|
||||
|
||||
确保 MySQL/Redis 容器与应用容器在同一 Docker 网络:
|
||||
|
||||
```bash
|
||||
# 创建共享网络(若不存在)
|
||||
docker network create edu-shared
|
||||
|
||||
# 将已有的 MySQL/Redis 加入网络(容器名按实际替换)
|
||||
docker network connect edu-shared edu-mysql # 或实际容器名
|
||||
docker network connect edu-shared edu-redis # 或实际容器名
|
||||
```
|
||||
|
||||
> 查看实际容器名:`docker ps --format "{{.Names}}" | grep -E "mysql|redis"`
|
||||
|
||||
### 2.4 初始化部署目录
|
||||
|
||||
```bash
|
||||
# 1. 创建部署目录
|
||||
sudo mkdir -p /opt/edu
|
||||
sudo chown -R $USER:$USER /opt/edu
|
||||
|
||||
# 2. 拷贝 compose 文件
|
||||
cd /path/to/Edu
|
||||
cp infra/docker-compose.deploy.yml /opt/edu/docker-compose.yml
|
||||
|
||||
# 3. 创建生产 .env(从模板)
|
||||
cp infra/deploy.env.example /opt/edu/.env
|
||||
vim /opt/edu/.env
|
||||
# 必须修改:
|
||||
# JWT_SECRET=<openssl rand -hex 32 生成的随机值>
|
||||
# DATABASE_URL=mysql://<user>:<pass>@<mysql容器名>:3306/<db>
|
||||
# REDIS_URL=redis://<redis容器名>:6379
|
||||
|
||||
# 4. 登录 Gitea Container Registry(Runner 用户)
|
||||
docker login git.eazygame.cn -u <user> -p <token>
|
||||
# token 在 Gitea → 设置 → 应用 → 生成新 token
|
||||
```
|
||||
|
||||
### 2.5 配置 Gitea Secrets
|
||||
|
||||
在 Gitea 仓库 → 设置 → Actions → Secrets,添加:
|
||||
|
||||
| Secret 名 | 用途 | 必填 |
|
||||
| ---------------------- | -------------------------------- | ---- |
|
||||
| `GITHUB_TOKEN` | Gitea 自动注入,推送镜像用 | 自动 |
|
||||
| `GITEA_REGISTRY_USER` | deploy.yml 登录 registry(可选) | 否 |
|
||||
| `GITEA_REGISTRY_TOKEN` | deploy.yml 登录 registry(可选) | 否 |
|
||||
|
||||
> `GITHUB_TOKEN` 由 Gitea Actions 自动注入,无需手动配置。`docker.yml` 用它推送镜像。
|
||||
> `deploy.yml` 中 Runner 已在服务器上,若已 `docker login` 过则无需 secrets。
|
||||
|
||||
---
|
||||
|
||||
## 3. 日常使用
|
||||
|
||||
### 3.1 开发 AI 提 PR
|
||||
|
||||
开发 AI 推送特性分支后创建 PR,自动触发 CI:
|
||||
|
||||
```
|
||||
PR 创建 → ci-ts.yml / ci-go.yml 运行
|
||||
├─ quality(lint + typecheck + test + build)
|
||||
├─ arch-scan(架构扫描)
|
||||
└─ docker-build(构建不推送)
|
||||
|
||||
CI 全绿 → 协调 AI 审核合并
|
||||
```
|
||||
|
||||
### 3.2 协调 AI 合并 PR
|
||||
|
||||
PR 合并到 main 后,自动触发:
|
||||
|
||||
```
|
||||
push main →
|
||||
├─ ci-*.yml(确保 main 稳定)
|
||||
└─ docker.yml → 构建并推送镜像到 git.eazygame.cn/xiner/edu/<service>:latest
|
||||
│
|
||||
▼
|
||||
deploy.yml → Runner 执行 docker compose pull && up -d
|
||||
│
|
||||
▼
|
||||
健康检查(轮询 /healthz)
|
||||
```
|
||||
|
||||
### 3.3 手动部署
|
||||
|
||||
在 Gitea → Actions → Deploy → Run workflow:
|
||||
|
||||
| 参数 | 说明 |
|
||||
| ----------- | --------------------------------------- |
|
||||
| `image_tag` | 部署指定 tag(默认 `latest`) |
|
||||
| `rollback` | 勾选则跳过 pull,用本地已有镜像重新启动 |
|
||||
|
||||
### 3.4 发布正式版本(tag)
|
||||
|
||||
```bash
|
||||
git tag -a v1.0.0 -m "P1 阶段首次发布"
|
||||
git push origin v1.0.0
|
||||
```
|
||||
|
||||
触发 `docker.yml` 推送 `v1.0.0` tag 镜像,再手动触发 deploy.yml 部署 `v1.0.0`。
|
||||
|
||||
---
|
||||
|
||||
## 4. 镜像管理
|
||||
|
||||
### 4.1 镜像地址
|
||||
|
||||
```
|
||||
git.eazygame.cn/xiner/edu/<service>:<tag>
|
||||
```
|
||||
|
||||
| service | 说明 |
|
||||
| ---------------- | --------------- |
|
||||
| `api-gateway` | Go 网关 |
|
||||
| `classes` | NestJS 班级服务 |
|
||||
| `teacher-portal` | Next.js 前端 |
|
||||
|
||||
### 4.2 Tag 策略
|
||||
|
||||
| tag | 来源 | 用途 |
|
||||
| ------------------ | ----------------- | ---------- |
|
||||
| `latest` | main 分支最新构建 | 默认部署 |
|
||||
| `main-<sha>` | main 分支每次构建 | 可追溯 |
|
||||
| `v<version>` | 打 `v*` tag 触发 | 正式版本 |
|
||||
| `v<major>.<minor>` | 打 `v*` tag 触发 | 大版本追踪 |
|
||||
|
||||
### 4.3 查看与清理镜像
|
||||
|
||||
在 Gitea → 用户设置 → Packages,可查看与管理所有镜像。
|
||||
|
||||
清理旧 tag(手动):
|
||||
|
||||
```bash
|
||||
# 列出所有 tag
|
||||
docker images git.eazygame.cn/xiner/edu/classes --format "{{.Tag}}"
|
||||
|
||||
# 删除本地旧镜像
|
||||
docker rmi git.eazygame.cn/xiner/edu/classes:main-abc1234
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 5. 部署验证
|
||||
|
||||
### 5.1 CI 自动验证
|
||||
|
||||
`deploy.yml` 部署后会自动轮询健康检查(最多 10 次,每次间隔 6 秒):
|
||||
|
||||
- `http://localhost:8080/healthz` — api-gateway
|
||||
- `http://localhost:3001/healthz` — classes
|
||||
- `http://localhost:3000/` — teacher-portal
|
||||
|
||||
失败时输出容器日志,方便排查。
|
||||
|
||||
### 5.2 手动验证
|
||||
|
||||
```bash
|
||||
# SSH 到服务器后
|
||||
cd /opt/edu
|
||||
|
||||
# 容器状态
|
||||
docker compose ps
|
||||
|
||||
# 健康检查
|
||||
curl http://localhost:8080/healthz # api-gateway
|
||||
curl http://localhost:3001/healthz # classes
|
||||
curl http://localhost:3000/ # teacher-portal
|
||||
|
||||
# 鉴权验证(生产模式 dev-token 应被拒绝)
|
||||
curl -H "Authorization: Bearer dev-token" http://localhost:8080/api/v1/classes
|
||||
# 预期:401 INVALID_TOKEN
|
||||
|
||||
# 查看日志
|
||||
docker compose logs -f api-gateway
|
||||
docker compose logs -f classes
|
||||
docker compose logs -f teacher-portal
|
||||
```
|
||||
|
||||
### 5.3 外部访问
|
||||
|
||||
若服务器有公网 IP(如 `1.2.3.4`):
|
||||
|
||||
- 前端:`http://1.2.3.4:3000`
|
||||
- API 网关:`http://1.2.3.4:8080`
|
||||
- 健康检查:`http://1.2.3.4:8080/healthz`
|
||||
|
||||
---
|
||||
|
||||
## 6. 回滚
|
||||
|
||||
### 6.1 CI 回滚(推荐)
|
||||
|
||||
在 Gitea → Actions → Deploy → Run workflow:
|
||||
|
||||
- `image_tag`:填入上一个稳定版本的 tag(如 `main-abc1234`)
|
||||
- `rollback`:勾选(跳过 pull,用本地已有镜像)
|
||||
|
||||
### 6.2 手动回滚
|
||||
|
||||
```bash
|
||||
cd /opt/edu
|
||||
|
||||
# 查看本地镜像
|
||||
docker images git.eazygame.cn/xiner/edu/classes --format "{{.Tag}}"
|
||||
|
||||
# 回滚到指定版本
|
||||
IMAGE_TAG=main-abc1234 docker compose up -d
|
||||
```
|
||||
|
||||
### 6.3 紧急回滚(Git revert)
|
||||
|
||||
若代码有问题:
|
||||
|
||||
```bash
|
||||
git revert <bad-commit>
|
||||
git push origin main
|
||||
# CI 自动重新构建部署
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 7. 常见问题
|
||||
|
||||
### 7.1 CI: pnpm install 失败
|
||||
|
||||
**症状**:`pnpm install --frozen-lockfile` 报错。
|
||||
|
||||
**排查**:
|
||||
|
||||
1. `pnpm-lock.yaml` 未提交:`git add pnpm-lock.yaml && git commit -m "chore(deps): 更新 lockfile"`
|
||||
2. Node 版本不匹配:确认 CI 用 Node 20,本地一致
|
||||
|
||||
### 7.2 CI: docker-build 失败
|
||||
|
||||
**症状**:`docker/build-push-action` 报错。
|
||||
|
||||
**排查**:
|
||||
|
||||
1. Dockerfile 语法错误:本地 `docker build` 验证
|
||||
2. 路径错误:确认 `context` 与 `file` 参数
|
||||
|
||||
### 7.3 CD: 镜像拉取失败
|
||||
|
||||
**症状**:`docker compose pull` 报 `unauthorized`。
|
||||
|
||||
**修复**:
|
||||
|
||||
```bash
|
||||
# 服务器上重新登录
|
||||
docker login git.eazygame.cn -u <user> -p <token>
|
||||
|
||||
# 或配置 Gitea secrets GITEA_REGISTRY_USER / GITEA_REGISTRY_TOKEN
|
||||
```
|
||||
|
||||
### 7.4 CD: 健康检查失败
|
||||
|
||||
**症状**:`deploy.yml` 健康检查 10 次后失败。
|
||||
|
||||
**排查**:
|
||||
|
||||
1. 查看日志:`cd /opt/edu && docker compose logs api-gateway`
|
||||
2. 常见原因:
|
||||
- `DATABASE_URL` 连不上 MySQL(检查容器名与网络)
|
||||
- `JWT_SECRET` 未配置
|
||||
- 端口被占用:`docker ps` 检查冲突
|
||||
|
||||
### 7.5 CD: 连不上 MySQL/Redis
|
||||
|
||||
**症状**:classes 容器报 `ECONNREFUSED edu-mysql:3306`。
|
||||
|
||||
**修复**:
|
||||
|
||||
```bash
|
||||
# 1. 确认 MySQL 容器名
|
||||
docker ps --format "{{.Names}}" | grep mysql
|
||||
|
||||
# 2. 确认 MySQL 在 edu-shared 网络
|
||||
docker network inspect edu-shared | grep -A5 Containers
|
||||
|
||||
# 3. 若不在,加入网络
|
||||
docker network connect edu-shared <实际mysql容器名>
|
||||
|
||||
# 4. 修改 /opt/edu/.env 的 DATABASE_URL
|
||||
# DATABASE_URL=mysql://edu:changeme@<实际容器名>:3306/next_edu_cloud
|
||||
```
|
||||
|
||||
### 7.6 Gitea Actions 未触发
|
||||
|
||||
**症状**:push 后 Actions 不运行。
|
||||
|
||||
**排查**:
|
||||
|
||||
1. 仓库设置 → Actions → 确认已启用
|
||||
2. Runner 在线:Gitea → 设置 → Actions → Runners
|
||||
3. workflow 文件在 `.github/workflows/` 目录(非 `.gitea/workflows/`)
|
||||
4. `on:` 触发条件匹配(paths 过滤)
|
||||
|
||||
### 7.7 workflow_run 不触发
|
||||
|
||||
**症状**:`docker.yml` 完成后 `deploy.yml` 不自动运行。
|
||||
|
||||
**原因**:Gitea Actions 对 `workflow_run` 触发支持可能不完整。
|
||||
|
||||
**解决**:
|
||||
|
||||
- 改用手动触发:Actions → Deploy → Run workflow
|
||||
- 或在 `docker.yml` 末尾加 job 调用 deploy(需要 `workflow_call`)
|
||||
|
||||
---
|
||||
|
||||
## 8. 排查命令速查
|
||||
|
||||
```bash
|
||||
# === 在服务器上 ===
|
||||
|
||||
# 查看所有 edu 容器
|
||||
docker compose -f /opt/edu/docker-compose.yml ps
|
||||
|
||||
# 查看实时日志
|
||||
docker compose -f /opt/edu/docker-compose.yml logs -f
|
||||
|
||||
# 重启单个服务
|
||||
docker compose -f /opt/edu/docker-compose.yml restart api-gateway
|
||||
|
||||
# 进入容器
|
||||
docker exec -it edu-api-gateway sh
|
||||
docker exec -it edu-classes sh
|
||||
|
||||
# 查看网络
|
||||
docker network inspect edu-shared
|
||||
|
||||
# 查看镜像列表
|
||||
docker images git.eazygame.cn/xiner/edu
|
||||
|
||||
# === 在 Gitea Web UI ===
|
||||
# 仓库 → Actions → 查看流水线运行记录
|
||||
# 用户设置 → Packages → 管理镜像
|
||||
# 仓库 → 设置 → Secrets → 管理 CI 密钥
|
||||
# 仓库 → 设置 → Actions → Runners → 查看 Runner 状态
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 9. 安全注意事项
|
||||
|
||||
1. **`.env` 文件不入库**:`.gitignore` 已忽略,仅存在于服务器 `/opt/edu/.env`
|
||||
2. **JWT_SECRET 强随机**:生产必须用 `openssl rand -hex 32` 生成
|
||||
3. **DEV_MODE=false**:docker-compose.deploy.yml 强制设为 `false`
|
||||
4. **Gitea Token 权限最小化**:仅给 `package:write` 和 `contents:read`
|
||||
5. **Runner 隔离**:deploy runner 仅跑在目标服务器,不暴露公网 SSH
|
||||
6. **镜像扫描**(P6):`docker.yml` 后续加 Trivy 扫描步骤
|
||||
|
||||
---
|
||||
|
||||
## 10. 相关文档
|
||||
|
||||
- [project_rules §15 CI/CD 规范](../../.trae/rules/project_rules.md)
|
||||
- [project_rules §14 多 AI 协作规范](../../.trae/rules/project_rules.md)
|
||||
- [本地启动手册](./local-dev-runbook.md)
|
||||
- [多 AI 协作指南](./multi-ai-collaboration.md)
|
||||
- [Git 工作流](./git-workflow.md)
|
||||
- [known-issues](../troubleshooting/known-issues.md)
|
||||
333
docs/standards/local-dev-runbook.md
Normal file
333
docs/standards/local-dev-runbook.md
Normal file
@@ -0,0 +1,333 @@
|
||||
# 本地启动手册(Local Dev Runbook)
|
||||
|
||||
> 版本:1.0
|
||||
> 日期:2026-07-08
|
||||
> 适用范围:Edu 微服务(P1 阶段:api-gateway + classes + teacher-portal)
|
||||
> 关联文档:[project_rules](../../.trae/rules/project_rules.md)、[004 架构影响地图](../architecture/004_architecture_impact_map.md)、[known-issues](../troubleshooting/known-issues.md)
|
||||
|
||||
---
|
||||
|
||||
## 1. 环境依赖
|
||||
|
||||
| 工具 | 版本要求 | 验证命令 | 说明 |
|
||||
| -------------- | -------- | ------------------------ | -------------------------- |
|
||||
| Node.js | ≥ 20 | `node -v` | LTS 版本 |
|
||||
| pnpm | ≥ 9 | `pnpm -v` | `corepack enable` 自动启用 |
|
||||
| Go | 1.22+ | `go version` | api-gateway 编译 |
|
||||
| Docker | 24+ | `docker version` | 基础设施容器化 |
|
||||
| Docker Compose | v2+ | `docker compose version` | 编排基础设施 |
|
||||
| Git | 2.30+ | `git --version` | husky hooks 需要 |
|
||||
|
||||
> Windows 用户:建议在 PowerShell 或 Git Bash 中执行。Go 工具链若不在 PATH,临时加入:`$env:Path = "C:\Program Files\Go\bin;" + $env:Path`
|
||||
|
||||
---
|
||||
|
||||
## 2. 服务端口分配
|
||||
|
||||
| 服务 | 端口 | 语言/框架 | 启动方式 | 健康检查 |
|
||||
| -------------- | ---- | ------------ | ------------------- | ----------------- |
|
||||
| MySQL | 3306 | Docker | `docker compose up` | `mysqladmin ping` |
|
||||
| Redis | 6379 | Docker | `docker compose up` | `redis-cli ping` |
|
||||
| api-gateway | 8080 | Go (Gin) | `go run main.go` | `GET /healthz` |
|
||||
| classes | 3001 | NestJS (TS) | `pnpm dev` | `GET /healthz` |
|
||||
| teacher-portal | 3000 | Next.js (TS) | `pnpm dev` | `GET /` |
|
||||
| iam | 3002 | NestJS | P2 阶段 | — |
|
||||
| teacher-bff | 3003 | NestJS | P2 阶段 | — |
|
||||
|
||||
> 端口冲突排查:`netstat -ano \| findstr :8080`(Windows)
|
||||
|
||||
---
|
||||
|
||||
## 3. 本地开发模式(DEV_MODE)
|
||||
|
||||
### 3.1 首次准备
|
||||
|
||||
```bash
|
||||
# 1. 克隆仓库
|
||||
git clone <repo-url> Edu
|
||||
cd Edu
|
||||
|
||||
# 2. 安装依赖
|
||||
pnpm install
|
||||
|
||||
# 3. 配置环境变量
|
||||
cp .env.example .env
|
||||
# 编辑 .env,确认:
|
||||
# DEV_MODE=true ← 本地联调用
|
||||
# JWT_SECRET=p1-dev-secret-change-in-production
|
||||
# DATABASE_URL=mysql://edu:changeme@localhost:3306/next_edu_cloud
|
||||
# REDIS_URL=redis://localhost:6379
|
||||
|
||||
# 4. 架构扫描(更新 arch.db)
|
||||
pnpm run arch:scan
|
||||
```
|
||||
|
||||
### 3.2 启动基础设施
|
||||
|
||||
```bash
|
||||
# 启动 MySQL + Redis(最小基础设施)
|
||||
docker compose -f infra/docker-compose.minimal.yml up -d
|
||||
|
||||
# 验证健康
|
||||
docker compose -f infra/docker-compose.minimal.yml ps
|
||||
# 状态应为 healthy
|
||||
|
||||
# (可选)国内镜像源加速:infra/docker-compose.minimal.override.yml 已配置 daocloud 镜像
|
||||
```
|
||||
|
||||
### 3.3 启动应用服务(三个终端)
|
||||
|
||||
#### 终端 1:api-gateway(:8080)
|
||||
|
||||
```bash
|
||||
cd services/api-gateway
|
||||
|
||||
# Windows:若 Go 不在 PATH,先设置
|
||||
$env:Path = "C:\Program Files\Go\bin;" + $env:Path
|
||||
|
||||
# 设置开发模式环境变量
|
||||
$env:DEV_MODE="true"
|
||||
$env:JWT_SECRET="p1-dev-secret-change-in-production"
|
||||
|
||||
# 启动
|
||||
go run main.go
|
||||
```
|
||||
|
||||
验证:`curl http://localhost:8080/healthz` → 200 OK
|
||||
|
||||
#### 终端 2:classes 服务(:3001)
|
||||
|
||||
```bash
|
||||
# 在项目根目录
|
||||
pnpm --filter @edu/classes-service dev
|
||||
```
|
||||
|
||||
验证:`curl http://localhost:3001/healthz` → 200 OK
|
||||
|
||||
> 注:classes 的 HealthModule 需注册到 AppModule,当前若返回 404 见 [known-issues]
|
||||
|
||||
#### 终端 3:teacher-portal(:3000)
|
||||
|
||||
```bash
|
||||
# 在项目根目录
|
||||
pnpm --filter @edu/teacher-portal dev
|
||||
```
|
||||
|
||||
验证:浏览器访问 `http://localhost:3000` → 班级管理页面
|
||||
|
||||
### 3.4 一键启动(可选)
|
||||
|
||||
```bash
|
||||
# 并行启动所有子包的 dev 脚本(含 api-gateway 需 go run,不会自动启动)
|
||||
pnpm dev
|
||||
```
|
||||
|
||||
> 注:`pnpm dev` 仅启动 pnpm workspace 内的 TS 服务。api-gateway 是 Go 服务,需单独 `go run`。
|
||||
|
||||
### 3.5 联调验证
|
||||
|
||||
```bash
|
||||
# 1. 直接访问 api-gateway(带 dev-token)
|
||||
curl -H "Authorization: Bearer dev-token" http://localhost:8080/api/v1/classes
|
||||
# 预期:200 OK + 班级列表 JSON
|
||||
|
||||
# 2. 通过 teacher-portal 代理访问
|
||||
curl -H "Authorization: Bearer dev-token" http://localhost:3000/api/v1/classes
|
||||
# 预期:200 OK(经 Next.js rewrites → api-gateway → classes)
|
||||
|
||||
# 3. 浏览器访问
|
||||
# http://localhost:3000 → 班级管理页面,自动加载班级列表
|
||||
```
|
||||
|
||||
### 3.6 dev-token 说明
|
||||
|
||||
- `DEV_MODE=true` 时,api-gateway 接受 `Authorization: Bearer dev-token`
|
||||
- 注入固定身份:`x-user-id: dev-user`,`x-user-roles: teacher,admin`
|
||||
- **仅限本地联调,生产环境必须 `DEV_MODE=false`**
|
||||
|
||||
---
|
||||
|
||||
## 4. 生产模式(Docker Compose)
|
||||
|
||||
### 4.1 构建镜像
|
||||
|
||||
```bash
|
||||
# 构建三个应用服务镜像
|
||||
docker compose -f infra/docker-compose.prod.yml build
|
||||
```
|
||||
|
||||
### 4.2 启动完整栈
|
||||
|
||||
```bash
|
||||
# 1. 先启动基础设施(MySQL + Redis)
|
||||
docker compose -f infra/docker-compose.minimal.yml up -d
|
||||
|
||||
# 2. 启动应用服务
|
||||
docker compose -f infra/docker-compose.prod.yml up -d
|
||||
|
||||
# 3. 查看状态
|
||||
docker compose -f infra/docker-compose.prod.yml ps
|
||||
# 所有服务应为 healthy
|
||||
|
||||
# 4. 查看日志
|
||||
docker compose -f infra/docker-compose.prod.yml logs -f api-gateway
|
||||
```
|
||||
|
||||
### 4.3 生产环境配置要点
|
||||
|
||||
| 配置项 | 生产值 | 说明 |
|
||||
| -------------- | ---------------- | ----------------------------------------------- |
|
||||
| `DEV_MODE` | `false` | docker-compose.prod.yml 强制设为 false |
|
||||
| `JWT_SECRET` | 强随机值 | 替换 `p1-dev-secret-change-in-production` |
|
||||
| `DATABASE_URL` | 生产数据库连接 | 容器内用 `host.docker.internal` 或 compose 网络 |
|
||||
| `NODE_ENV` | `production` | teacher-portal 启用生产优化 |
|
||||
| `LOG_LEVEL` | `info` 或 `warn` | 生产日志级别 |
|
||||
|
||||
### 4.4 生产验证
|
||||
|
||||
```bash
|
||||
# 健康检查
|
||||
curl http://localhost:8080/healthz # api-gateway
|
||||
curl http://localhost:3001/healthz # classes
|
||||
curl http://localhost:3000/ # teacher-portal
|
||||
|
||||
# 鉴权验证(生产模式 dev-token 应被拒绝)
|
||||
curl -H "Authorization: Bearer dev-token" http://localhost:8080/api/v1/classes
|
||||
# 预期:401 INVALID_TOKEN
|
||||
|
||||
# 真实 JWT 访问(需 IAM 签发)
|
||||
curl -H "Authorization: Bearer <real-jwt>" http://localhost:8080/api/v1/classes
|
||||
# 预期:200 OK
|
||||
```
|
||||
|
||||
### 4.5 停止与清理
|
||||
|
||||
```bash
|
||||
# 停止应用服务
|
||||
docker compose -f infra/docker-compose.prod.yml down
|
||||
|
||||
# 停止基础设施
|
||||
docker compose -f infra/docker-compose.minimal.yml down
|
||||
|
||||
# 清理数据卷(谨慎!会删除数据库数据)
|
||||
docker compose -f infra/docker-compose.minimal.yml down -v
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 5. 单独构建与运行(裸机生产)
|
||||
|
||||
### 5.1 api-gateway
|
||||
|
||||
```bash
|
||||
cd services/api-gateway
|
||||
|
||||
# 编译
|
||||
go build -o bin/api-gateway ./main.go
|
||||
|
||||
# 运行(生产环境变量)
|
||||
DEV_MODE=false JWT_SECRET=<your-secret> ./bin/api-gateway
|
||||
```
|
||||
|
||||
### 5.2 classes 服务
|
||||
|
||||
```bash
|
||||
# 构建
|
||||
pnpm --filter @edu/classes-service build
|
||||
|
||||
# 运行(从 dist/ 启动)
|
||||
NODE_ENV=production PORT=3001 \
|
||||
DATABASE_URL=mysql://edu:changeme@localhost:3306/next_edu_cloud \
|
||||
REDIS_URL=redis://localhost:6379 \
|
||||
node services/classes/dist/main.js
|
||||
```
|
||||
|
||||
### 5.3 teacher-portal
|
||||
|
||||
```bash
|
||||
# 构建
|
||||
pnpm --filter @edu/teacher-portal build
|
||||
|
||||
# 运行
|
||||
NODE_ENV=production PORT=3000 \
|
||||
API_GATEWAY_URL=http://localhost:8080 \
|
||||
node apps/teacher-portal/node_modules/.bin/next start -p 3000
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 6. 常见问题
|
||||
|
||||
### 6.1 ERR_TOO_MANY_REDIRECTS
|
||||
|
||||
**症状**:浏览器访问 `:3000/api/v1/classes` 无限重定向。
|
||||
|
||||
**根因**:Gin 默认 `RedirectTrailingSlash=true`,`/classes` → 301 → `/classes/`,Next.js rewrites 代理形成循环。
|
||||
|
||||
**修复**:`main.go` 已设 `r.RedirectTrailingSlash = false`,并同时注册无尾斜杠与通配符路由。
|
||||
|
||||
### 6.2 Docker Hub 拉取超时
|
||||
|
||||
**症状**:`docker compose up` 时 MySQL/Redis 镜像拉取超时。
|
||||
|
||||
**修复**:`infra/docker-compose.minimal.override.yml` 已配置 daocloud 镜像源,docker compose 会自动加载。
|
||||
|
||||
### 6.3 classes DI 注入失败
|
||||
|
||||
**症状**:`TypeError: Cannot read properties of undefined (reading 'create')`。
|
||||
|
||||
**根因**:NestJS ESM 模式下 `emitDecoratorMetadata` 不工作。
|
||||
|
||||
**修复**:`ClassesService` 构造函数已加 `@Inject(ClassesRepository)` 显式指定 token。
|
||||
|
||||
### 6.4 ESM import 缺 .js 后缀
|
||||
|
||||
**症状**:`error TS2307: Cannot find module './health.controller'`。
|
||||
|
||||
**修复**:ESM 模式下相对 import 必须带 `.js` 后缀(详见 project_rules §3.4)。
|
||||
|
||||
### 6.5 JWT 401 INVALID_TOKEN
|
||||
|
||||
**症状**:带 `dev-token` 仍返回 401。
|
||||
|
||||
**排查**:
|
||||
|
||||
1. 确认 `DEV_MODE=true` 环境变量已设置(api-gateway 进程)
|
||||
2. 确认请求头格式:`Authorization: Bearer dev-token`(注意 Bearer 后空格)
|
||||
3. 生产模式(`DEV_MODE=false`)下 dev-token 被拒绝是预期行为
|
||||
|
||||
### 6.6 Go 工具链不在 PATH
|
||||
|
||||
**症状**:Git Bash 或 PowerShell 中 `go: command not found`。
|
||||
|
||||
**修复**:
|
||||
|
||||
```powershell
|
||||
$env:Path = "C:\Program Files\Go\bin;" + $env:Path
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 7. 数据库初始化
|
||||
|
||||
classes 服务使用 Drizzle ORM,首次运行需建表:
|
||||
|
||||
```bash
|
||||
# 生成迁移
|
||||
pnpm --filter @edu/classes-service drizzle-kit generate
|
||||
|
||||
# 执行迁移
|
||||
pnpm --filter @edu/classes-service drizzle-kit migrate
|
||||
```
|
||||
|
||||
> 若 `drizzle-kit` 未配置,可手动执行 `services/classes/src/db/schema.sql`(如存在)。
|
||||
|
||||
---
|
||||
|
||||
## 8. 相关文档
|
||||
|
||||
- [项目规则](../../.trae/rules/project_rules.md) — 强制约束
|
||||
- [004 架构影响地图](../architecture/004_architecture_impact_map.md) — 服务清单与调用关系
|
||||
- [known-issues](../troubleshooting/known-issues.md) — 已知问题速查
|
||||
- [Git 工作流](./git-workflow.md) — 提交与 PR 规范
|
||||
- [多 AI 协作指南](./multi-ai-collaboration.md) — 多 Agent 并行开发流程
|
||||
735
docs/standards/multi-ai-collaboration.md
Normal file
735
docs/standards/multi-ai-collaboration.md
Normal file
@@ -0,0 +1,735 @@
|
||||
# 多 AI 协作开发指南(Multi-AI Collaboration Guide)
|
||||
|
||||
> 版本:1.0
|
||||
> 日期:2026-07-08
|
||||
> 适用范围:Edu 微服务项目多 Agent 并行开发
|
||||
> 关联文档:[Git 工作流](./git-workflow.md)、[项目规则](../../.trae/rules/project_rules.md)、[004 架构影响地图](../architecture/004_architecture_impact_map.md)、[CODEOWNERS](../../.github/CODEOWNERS)
|
||||
|
||||
---
|
||||
|
||||
## 1. 总体模式
|
||||
|
||||
### 1.1 角色定义
|
||||
|
||||
| 角色 | 职责 | 数量 | 备注 |
|
||||
| ----------------------------- | -------------------------------------------- | ---- | ------------------- |
|
||||
| **协调 AI(Coordinator AI)** | PR 审核、合并、冲突仲裁、分支管理、发布 | 1 | 不直接写业务代码 |
|
||||
| **开发 AI(Dev AI)** | 按模块分工,写代码、提 PR | N | 每个负责 1-2 个模块 |
|
||||
| **人类决策者(Human)** | 架构决策、Breaking Change 审批、最终发布确认 | 1 | 只在关键节点介入 |
|
||||
|
||||
### 1.2 工作流总览
|
||||
|
||||
```
|
||||
开发 AI-A (classes模块) 开发 AI-B (iam模块) 协调 AI
|
||||
│ │ │
|
||||
│ 1.拉取最新 main │ 1.拉取最新 main │
|
||||
│ 2.创建特性分支 │ 2.创建特性分支 │
|
||||
│ 3.开发+提交 │ 3.开发+提交 │
|
||||
│ 4.推送+创建 PR │ 4.推送+创建 PR │
|
||||
│──────────────────────────────────────────────────────→│
|
||||
│ │ │ 5.CI校验
|
||||
│ │ │ 6.代码审核
|
||||
│ │ │ 7.合并到 main
|
||||
│ │ │ 8.通知开发 AI 同步
|
||||
│←─────────────────────────────────────────────────────│
|
||||
│ 9.拉取最新 main │ │
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 2. 模块分工矩阵
|
||||
|
||||
### 2.1 模块清单与 AI 分配
|
||||
|
||||
每个 AI 负责一个"限界上下文",避免跨模块修改导致冲突。
|
||||
|
||||
| 模块 | scope | 路径 | 负责阶段 | 建议 AI 数 |
|
||||
| -------------- | ---------------- | ------------------------- | -------- | ------------ |
|
||||
| api-gateway | `api-gateway` | `services/api-gateway/` | P1 | 1 |
|
||||
| classes | `classes` | `services/classes/` | P1 | 1 |
|
||||
| teacher-portal | `teacher-portal` | `apps/teacher-portal/` | P1-P2 | 1 |
|
||||
| iam | `iam` | `services/iam/` | P2 | 1 |
|
||||
| teacher-bff | `teacher-bff` | `services/teacher-bff/` | P2 | 1 |
|
||||
| core-edu | `core-edu` | `services/core-edu/` | P3 | 1 |
|
||||
| content | `content` | `services/content/` | P4 | 1 |
|
||||
| data-ana | `data-ana` | `services/data-ana/` | P4 | 1 |
|
||||
| msg | `msg` | `services/msg/` | P5 | 1 |
|
||||
| ai | `ai` | `services/ai/` | P5 | 1 |
|
||||
| push-gateway | `push-gateway` | `services/push-gateway/` | P5 | 1 |
|
||||
| shared-proto | `shared-proto` | `packages/shared-proto/` | 跨阶段 | 协调 AI 维护 |
|
||||
| shared-tokens | `shared-tokens` | `packages/shared-tokens/` | 跨阶段 | 协调 AI 维护 |
|
||||
| infra | `infra` | `infra/` | 跨阶段 | 1 (SRE AI) |
|
||||
| docs | `docs` | `docs/` | 跨阶段 | 协调 AI 维护 |
|
||||
|
||||
### 2.2 分工原则
|
||||
|
||||
1. **单一负责制**:每个模块只有一个 AI 负责,避免并行修改同一文件
|
||||
2. **契约集中管理**:`shared-proto` 由协调 AI 维护,开发 AI 只读引用
|
||||
3. **跨模块变更拆分**:需要修改多个模块时,拆成多个 PR,按依赖顺序合并
|
||||
4. **基础设施独立**:`infra/` 由 SRE AI 专门负责,业务 AI 不直接修改
|
||||
|
||||
---
|
||||
|
||||
## 3. 分支策略
|
||||
|
||||
### 3.1 分支命名规范
|
||||
|
||||
```
|
||||
<type>/<scope>-<task-id>-<ai-id>
|
||||
```
|
||||
|
||||
| 字段 | 说明 | 示例 |
|
||||
| --------- | ------------------------------------------- | ---------------- |
|
||||
| `type` | feat / fix / refactor / docs / chore / test | `feat` |
|
||||
| `scope` | 模块名(见 §2.1) | `classes` |
|
||||
| `task-id` | 任务简短描述(kebab-case) | `add-pagination` |
|
||||
| `ai-id` | AI 标识符(避免多 AI 同名冲突) | `ai01` |
|
||||
|
||||
**完整示例**:
|
||||
|
||||
- `feat/classes-add-pagination-ai01`
|
||||
- `fix/api-gateway-redirect-loop-ai02`
|
||||
- `docs/git-workflow-update-coord`
|
||||
|
||||
### 3.2 分支生命周期
|
||||
|
||||
1. **创建**:从最新 `main` 拉取:`git checkout main && git pull && git checkout -b feat/classes-xxx-ai01`
|
||||
2. **开发**:在特性分支上提交(遵循 Conventional Commits)
|
||||
3. **推送**:`git push -u origin feat/classes-xxx-ai01`
|
||||
4. **合并后删除**:PR 合并后,删除本地与远程特性分支
|
||||
|
||||
> 特性分支寿命 ≤ 3 天(见 git-workflow.md §1)。超期需 rebase 最新 main。
|
||||
|
||||
### 3.3 多 AI 分支隔离规则
|
||||
|
||||
- **禁止多 AI 共用同一分支**
|
||||
- **禁止直接 push 到 `main`**(由协调 AI 通过 PR 合并)
|
||||
- **跨模块变更**:各自在模块分支开发,最后由协调 AI 按依赖顺序合并
|
||||
|
||||
---
|
||||
|
||||
## 4. 推送与提交流程
|
||||
|
||||
### 4.1 提交规范(Conventional Commits)
|
||||
|
||||
```bash
|
||||
git commit -m "feat(classes): 新增班级分页查询"
|
||||
git commit -m "fix(api-gateway): 修复尾斜杠重定向循环"
|
||||
git commit -m "docs(standards): 更新多AI协作文档"
|
||||
```
|
||||
|
||||
**scope 必须匹配 `.commitlintrc.js` 的 scope-enum**(26 项,见 git-workflow.md §3.3)。
|
||||
|
||||
### 4.2 标准推送流程
|
||||
|
||||
开发 AI 执行:
|
||||
|
||||
```bash
|
||||
# 1. 确认在正确的特性分支
|
||||
git branch --show-current
|
||||
# 输出:feat/classes-add-pagination-ai01
|
||||
|
||||
# 2. 拉取最新 main(rebase 保持线性)
|
||||
git fetch origin
|
||||
git rebase origin/main
|
||||
|
||||
# 3. 解决冲突(如有)
|
||||
# 编辑冲突文件 → git add → git rebase --continue
|
||||
|
||||
# 4. 本地校验(强制,见 project_rules §7)
|
||||
pnpm run lint # TS 服务
|
||||
pnpm run typecheck # TS 服务
|
||||
cd services/api-gateway && go vet ./... && go build ./... # Go 服务
|
||||
|
||||
# 5. 推送
|
||||
git push -u origin feat/classes-add-pagination-ai01
|
||||
|
||||
# 6. 创建 PR(见 §5)
|
||||
```
|
||||
|
||||
### 4.3 多 AI 推送冲突处理
|
||||
|
||||
当多个 AI 的 PR 修改了同一文件(如 `package.json`、`004_architecture_impact_map.md`):
|
||||
|
||||
1. **先合并的 PR 正常合并**
|
||||
2. **后合并的 PR 需 rebase**:
|
||||
```bash
|
||||
git fetch origin
|
||||
git rebase origin/main
|
||||
# 解决冲突
|
||||
git push --force-with-lease # 安全强推(仅自己的分支)
|
||||
```
|
||||
3. **协调 AI 仲裁**:若冲突涉及架构决策,由协调 AI 决定保留方案
|
||||
|
||||
> **禁止 `git push --force` 到 main 或他人分支**。只允许 `--force-with-lease` 到自己的特性分支。
|
||||
|
||||
---
|
||||
|
||||
## 5. PR 流程
|
||||
|
||||
### 5.1 创建 PR
|
||||
|
||||
推送后,开发 AI 通过 Git 平台(GitHub/Gitea)创建 PR:
|
||||
|
||||
```bash
|
||||
# 使用 gh CLI(GitHub)
|
||||
gh pr create \
|
||||
--title "feat(classes): 新增班级分页查询" \
|
||||
--body "## 变更说明
|
||||
- 新增 GET /api/v1/classes?page=&size= 分页参数
|
||||
- 响应增加 total/hasMore 字段
|
||||
|
||||
## 变更类型
|
||||
- [x] feat(新功能)
|
||||
|
||||
## 影响范围
|
||||
- [x] classes 服务
|
||||
|
||||
## 测试
|
||||
- [x] 单元测试通过
|
||||
- [x] 本地联调验证
|
||||
|
||||
## 校验
|
||||
- [x] pnpm run lint
|
||||
- [x] pnpm run typecheck
|
||||
" \
|
||||
--base main \
|
||||
--head feat/classes-add-pagination-ai01
|
||||
```
|
||||
|
||||
### 5.2 PR 标题规范
|
||||
|
||||
PR 标题必须与首个 commit 的 subject 一致(Squash Merge 时自动取首个 commit):
|
||||
|
||||
```
|
||||
<type>(<scope>): <subject>
|
||||
```
|
||||
|
||||
**示例**:
|
||||
|
||||
- `feat(classes): 新增班级分页查询`
|
||||
- `fix(api-gateway): 修复尾斜杠重定向循环`
|
||||
|
||||
### 5.3 PR 模板
|
||||
|
||||
使用 `.github/pull_request_template.md`(已存在)。必填项:
|
||||
|
||||
- 变更说明
|
||||
- 变更类型
|
||||
- 影响范围
|
||||
- 测试情况
|
||||
- 按语言校验结果
|
||||
- 文档同步(arch:scan、CODEOWNERS)
|
||||
|
||||
### 5.4 PR 标注 AI 身份
|
||||
|
||||
在 PR 描述末尾追加(用于统计与审计):
|
||||
|
||||
```markdown
|
||||
---
|
||||
|
||||
**AI Agent**: ai01 (classes-module)
|
||||
**Branch**: feat/classes-add-pagination-ai01
|
||||
**Coordinator**: coord-ai
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 6. 代码审核与合并
|
||||
|
||||
### 6.1 Reviewer 分配
|
||||
|
||||
由 `.github/CODEOWNERS` 自动分配:
|
||||
|
||||
| 模块 | 自动 Reviewer | 实际映射 |
|
||||
| ------------ | ----------------------- | -------- |
|
||||
| classes | `@edu-platform/classes` | 协调 AI |
|
||||
| api-gateway | `@edu-platform/gateway` | 协调 AI |
|
||||
| shared-proto | `@edu-platform/arch` | 协调 AI |
|
||||
| infra | `@edu-platform/sre` | SRE AI |
|
||||
|
||||
> 当前 `@edu-platform/*` 为占位符。多 AI 场景下,协调 AI 账号加入对应 team 即可自动收到 review 请求。
|
||||
|
||||
### 6.2 审核 checklist
|
||||
|
||||
协调 AI 审核 PR 时检查:
|
||||
|
||||
- [ ] **Commit 格式**:符合 Conventional Commits
|
||||
- [ ] **Scope 正确**:scope 与修改的模块一致
|
||||
- [ ] **架构合规**:无跨层依赖、无直接访问他人 DB
|
||||
- [ ] **权限校验**:Controller 有 `@RequirePermission()`
|
||||
- [ ] **设计令牌**:无硬编码颜色/字体(project_rules §3.10)
|
||||
- [ ] **ESM 规则**:相对 import 带 `.js` 后缀(TS 服务)
|
||||
- [ ] **契约同步**:proto 变更已同步 shared-proto
|
||||
- [ ] **文档同步**:arch.db 已更新、004 已同步(如涉及)
|
||||
- [ ] **CI 通过**:lint / typecheck / go vet / ruff 全绿
|
||||
- [ ] **测试覆盖**:关键逻辑有单测
|
||||
|
||||
### 6.3 合并策略
|
||||
|
||||
| 策略 | 适用场景 | 操作 |
|
||||
| ------------------------ | ------------------------------------ | ---------------------- |
|
||||
| **Squash Merge**(默认) | 单 commit 或多个小 commit 合并为一个 | `gh pr merge --squash` |
|
||||
| **Rebase Merge** | 多个有意义的 commit 需保留历史 | `gh pr merge --rebase` |
|
||||
| **Merge Commit**(禁止) | — | 不允许 |
|
||||
|
||||
> 项目规则 git-workflow.md §1.4:**禁止 Merge Commit**,保持 main 历史线性。
|
||||
|
||||
### 6.4 合并流程(协调 AI 执行)
|
||||
|
||||
```bash
|
||||
# 1. 确认 CI 通过
|
||||
gh pr checks <PR-NUMBER>
|
||||
|
||||
# 2. 确认 review 通过(CODEOWNERS 已 approve)
|
||||
gh pr view <PR-NUMBER> --json reviews
|
||||
|
||||
# 3. Squash 合并(删除源分支)
|
||||
gh pr merge <PR-NUMBER> --squash --delete-branch
|
||||
|
||||
# 4. 通知开发 AI 同步本地 main
|
||||
# (通过 commit message 或 issue 评论)
|
||||
```
|
||||
|
||||
### 6.5 合并后同步
|
||||
|
||||
开发 AI 收到合并通知后:
|
||||
|
||||
```bash
|
||||
# 切回 main
|
||||
git checkout main
|
||||
|
||||
# 拉取最新(含已合并的 PR)
|
||||
git pull origin main
|
||||
|
||||
# 删除本地特性分支
|
||||
git branch -d feat/classes-add-pagination-ai01
|
||||
|
||||
# 开始下一个任务:从最新 main 拉新分支
|
||||
git checkout -b feat/classes-next-task-ai01
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 7. 多 AI 并行开发流程
|
||||
|
||||
### 7.1 启动并行任务
|
||||
|
||||
人类决策者分配任务给多个 AI:
|
||||
|
||||
```
|
||||
任务 1:classes 服务新增分页查询 → 开发 AI-A (ai01)
|
||||
任务 2:api-gateway 新增限流规则 → 开发 AI-B (ai02)
|
||||
任务 3:teacher-portal 优化班级列表 → 开发 AI-C (ai03)
|
||||
```
|
||||
|
||||
### 7.2 并行执行
|
||||
|
||||
每个 AI 独立工作,互不干扰:
|
||||
|
||||
```bash
|
||||
# AI-A
|
||||
git checkout -b feat/classes-add-pagination-ai01
|
||||
# ... 开发 ...
|
||||
git push -u origin feat/classes-add-pagination-ai01
|
||||
gh pr create --title "feat(classes): 新增班级分页查询" ...
|
||||
|
||||
# AI-B(同时)
|
||||
git checkout -b feat/api-gateway-rate-limit-ai02
|
||||
# ... 开发 ...
|
||||
git push -u origin feat/api-gateway-rate-limit-ai02
|
||||
gh pr create --title "feat(api-gateway): 新增 IP 级限流规则" ...
|
||||
|
||||
# AI-C(同时)
|
||||
git checkout -b feat/teacher-portal-class-list-ux-ai03
|
||||
# ... 开发 ...
|
||||
```
|
||||
|
||||
### 7.3 顺序合并(有依赖时)
|
||||
|
||||
若 PR 间有依赖(如 AI-C 依赖 AI-A 的接口),协调 AI 按顺序合并:
|
||||
|
||||
1. 合并 AI-A 的 PR(classes 分页接口)
|
||||
2. AI-C rebase 最新 main,解决冲突,更新 PR
|
||||
3. 合并 AI-C 的 PR(前端调用新接口)
|
||||
|
||||
### 7.4 冲突预防
|
||||
|
||||
1. **契约先行**:proto 变更先合并,再合并依赖它的服务
|
||||
2. **通知机制**:修改 `shared-proto/` 或 `004_architecture_impact_map.md` 时,在 PR 描述 @ 所有关联模块 AI
|
||||
3. **频繁同步**:每天至少 `git pull origin main` 一次,避免长期分叉
|
||||
|
||||
---
|
||||
|
||||
## 8. 跨模块变更流程
|
||||
|
||||
### 8.1 变更拆分原则
|
||||
|
||||
修改涉及多模块时,按依赖顺序拆成多个 PR:
|
||||
|
||||
```
|
||||
原始任务:classes 服务新增"班级导入"功能
|
||||
涉及:
|
||||
1. shared-proto 新增 ImportClassesRequest message
|
||||
2. classes 服务实现 import 接口
|
||||
3. api-gateway 路由 /api/v1/classes/import
|
||||
4. teacher-portal 前端上传按钮
|
||||
```
|
||||
|
||||
### 8.2 拆分与合并顺序
|
||||
|
||||
| 顺序 | PR | scope | 负责人 | 依赖 |
|
||||
| ---- | ----------------------------------------------- | -------------- | --------- | ---- |
|
||||
| 1 | `feat(shared-proto): 新增 ImportClassesRequest` | shared-proto | 协调 AI | 无 |
|
||||
| 2 | `feat(classes): 实现班级批量导入` | classes | 开发 AI-A | PR-1 |
|
||||
| 3 | `feat(api-gateway): 新增 import 路由` | api-gateway | 开发 AI-B | PR-2 |
|
||||
| 4 | `feat(teacher-portal): 新增班级导入按钮` | teacher-portal | 开发 AI-C | PR-3 |
|
||||
|
||||
协调 AI 按顺序合并,每合并一个,后续 PR 的开发 AI rebase 最新 main。
|
||||
|
||||
### 8.3 Breaking Change 流程
|
||||
|
||||
涉及 Breaking Change(proto 字段删除、API 签名变更):
|
||||
|
||||
1. **人类决策者审批**:在 issue 中讨论,获批准后才开发
|
||||
2. **版本号升级**:proto 加 `v2` 后缀,服务同时支持 v1/v2 过渡期
|
||||
3. **迁移文档**:更新 `MIGRATION_GUIDE.md`
|
||||
4. **协调 AI 通知所有相关 AI**:在 PR 描述中列出影响范围
|
||||
|
||||
---
|
||||
|
||||
## 9. 完整工作流示例
|
||||
|
||||
### 9.1 场景:开发 AI-A 为 classes 新增分页查询
|
||||
|
||||
```bash
|
||||
# ============ 开发 AI-A ============
|
||||
|
||||
# 1. 同步 main
|
||||
git checkout main
|
||||
git pull origin main
|
||||
|
||||
# 2. 创建特性分支
|
||||
git checkout -b feat/classes-add-pagination-ai01
|
||||
|
||||
# 3. 开发
|
||||
# 编辑 services/classes/src/classes/classes.controller.ts
|
||||
# 编辑 services/classes/src/classes/classes.service.ts
|
||||
# 编辑 services/classes/src/classes/classes.repository.ts
|
||||
|
||||
# 4. 本地校验
|
||||
pnpm --filter @edu/classes-service lint
|
||||
pnpm --filter @edu/classes-service typecheck
|
||||
pnpm --filter @edu/classes-service test
|
||||
|
||||
# 5. 架构扫描(若新增了导出函数/路由)
|
||||
pnpm run arch:scan
|
||||
|
||||
# 6. 提交
|
||||
git add services/classes/
|
||||
git commit -m "feat(classes): 新增班级分页查询
|
||||
|
||||
- GET /api/v1/classes 支持 page、size 参数
|
||||
- 响应增加 total、hasMore 字段
|
||||
- 单测覆盖分页逻辑"
|
||||
|
||||
# 7. 推送
|
||||
git push -u origin feat/classes-add-pagination-ai01
|
||||
|
||||
# 8. 创建 PR
|
||||
gh pr create \
|
||||
--title "feat(classes): 新增班级分页查询" \
|
||||
--body "..." \
|
||||
--base main
|
||||
|
||||
# ============ 协调 AI ============
|
||||
|
||||
# 9. 收到 PR 通知,审核
|
||||
gh pr view <PR-NUMBER>
|
||||
gh pr checks <PR-NUMBER>
|
||||
|
||||
# 10. 代码审核(见 §6.2 checklist)
|
||||
# 若需修改,评论要求开发 AI-A 修改
|
||||
|
||||
# 11. 合并
|
||||
gh pr merge <PR-NUMBER> --squash --delete-branch
|
||||
|
||||
# 12. 通知开发 AI-A
|
||||
# 评论:"已合并,请同步本地 main"
|
||||
|
||||
# ============ 开发 AI-A ============
|
||||
|
||||
# 13. 同步
|
||||
git checkout main
|
||||
git pull origin main
|
||||
git branch -d feat/classes-add-pagination-ai01
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 10. 常见问题与解决方案
|
||||
|
||||
### 10.1 多 AI 修改同一文件冲突
|
||||
|
||||
**场景**:AI-A 和 AI-B 都修改了 `package.json` 添加依赖。
|
||||
|
||||
**解决**:
|
||||
|
||||
1. AI-A 的 PR 先合并
|
||||
2. AI-B 执行:
|
||||
```bash
|
||||
git fetch origin
|
||||
git rebase origin/main
|
||||
# 解决 package.json 冲突(保留两边依赖)
|
||||
git add package.json
|
||||
git rebase --continue
|
||||
git push --force-with-lease
|
||||
```
|
||||
3. 协调 AI 重新审核
|
||||
|
||||
### 10.2 CI 失败但本地通过
|
||||
|
||||
**场景**:开发 AI 本地 lint 通过,CI 报错。
|
||||
|
||||
**排查**:
|
||||
|
||||
1. 检查 Node/Go/pnpm 版本是否一致(见 local-dev-runbook §1)
|
||||
2. 检查 `pnpm-lock.yaml` 是否最新(`pnpm install --frozen-lockfile`)
|
||||
3. 检查环境变量是否齐全(`.env` vs CI secrets)
|
||||
|
||||
**修复后**:
|
||||
|
||||
```bash
|
||||
git add <fixed-files>
|
||||
git commit -m "fix(classes): 修复 CI lint 失败"
|
||||
git push
|
||||
```
|
||||
|
||||
### 10.3 husky pre-commit hook 失败
|
||||
|
||||
**场景**:commit 时 hook 报错(lint-staged、commitlint)。
|
||||
|
||||
**排查**:
|
||||
|
||||
1. commitlint:检查 commit message 格式(type 小写、scope 在 enum 内、subject 不大写开头)
|
||||
2. lint-staged:检查暂存文件是否通过 ESLint/gofmt
|
||||
3. husky 9 Windows bug:若 pre-push hook "Bad file descriptor",见 known-issues
|
||||
|
||||
**修复**:修正后重新 `git commit`,不要用 `--no-verify` 跳过。
|
||||
|
||||
### 10.4 proto 变更导致下游编译失败
|
||||
|
||||
**场景**:shared-proto 合并后,classes 服务 `buf generate` 报错。
|
||||
|
||||
**解决**:
|
||||
|
||||
1. 协调 AI 合并 proto PR 前,先在 PR 评论中通知所有依赖服务的 AI
|
||||
2. 下游 AI 在自己的分支 rebase 最新 main 后重新 `buf generate`
|
||||
3. 更新生成代码并提交:
|
||||
```bash
|
||||
git add packages/shared-proto/gen/
|
||||
git commit -m "chore(shared-proto): 重新生成 TS 类型"
|
||||
```
|
||||
|
||||
### 10.5 分支长期未合并(超 3 天)
|
||||
|
||||
**场景**:特性分支超过 3 天未合并。
|
||||
|
||||
**处理**:
|
||||
|
||||
1. rebase 最新 main:`git rebase origin/main`
|
||||
2. 解决冲突后 `git push --force-with-lease`
|
||||
3. 若仍无法合并,协调 AI 介入评估是否拆分 PR
|
||||
|
||||
### 10.6 误推到 main
|
||||
|
||||
**场景**:开发 AI 误将 commit 推到 main。
|
||||
|
||||
**修复**(协调 AI 执行):
|
||||
|
||||
```bash
|
||||
# 1. 确认误推的 commit
|
||||
git log origin/main -5
|
||||
|
||||
# 2. 回退(保留误推 commit 到临时分支)
|
||||
git checkout origin/main -b temp/backup-mispush
|
||||
git checkout main
|
||||
git reset --hard origin/main~1 # 回退 1 个 commit
|
||||
|
||||
# 3. 强推(仅协调 AI 有权限)
|
||||
git push --force-with-lease origin main
|
||||
|
||||
# 4. 通知开发 AI 从 temp 分支重新提 PR
|
||||
```
|
||||
|
||||
> **禁止开发 AI 自行 force push main**。只有协调 AI 有权操作。
|
||||
|
||||
### 10.7 AI 身份冲突(同名分支)
|
||||
|
||||
**场景**:两个 AI 都叫 "ai01",创建了同名分支。
|
||||
|
||||
**预防**:每个 AI 启动前分配唯一 `ai-id`(如 `ai01`、`ai02`、`coord`)。
|
||||
|
||||
**修复**:后启动的 AI 重命名分支:
|
||||
|
||||
```bash
|
||||
git branch -m feat/classes-xxx-ai01 feat/classes-xxx-ai01b
|
||||
git push -u origin feat/classes-xxx-ai01b
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 11. AI 协作通信约定
|
||||
|
||||
### 11.1 通信渠道
|
||||
|
||||
| 场景 | 方式 |
|
||||
| -------- | ----------------------------- |
|
||||
| PR 审核 | PR 评论(`@ai01 请修改 xxx`) |
|
||||
| 任务分配 | Issue(`@ai01 负责实现 xxx`) |
|
||||
| 紧急通知 | PR 评论 + @ 相关人员 |
|
||||
| 架构讨论 | Issue 标签 `discussion` |
|
||||
|
||||
### 11.2 PR 评论规范
|
||||
|
||||
协调 AI 审核评论格式:
|
||||
|
||||
```
|
||||
## 审核结果:需修改 / 通过 / 拒绝
|
||||
|
||||
### 需修改项
|
||||
1. [classes.controller.ts:42] 缺少 @RequirePermission() 装饰器
|
||||
2. [classes.service.ts:88] 返回值未标注 Promise<T>
|
||||
|
||||
### 建议
|
||||
- 考虑抽取分页逻辑到 shared-ts
|
||||
|
||||
---
|
||||
协调 AI(coord)
|
||||
```
|
||||
|
||||
### 11.3 AI 工作日志
|
||||
|
||||
每个 AI 完成任务后,在 `docs/troubleshooting/known-issues.md` "工作经验日志"区追加(project_rules §9.3):
|
||||
|
||||
```markdown
|
||||
| 日期 | 模块 | 做了什么 + 学到什么 |
|
||||
| ---------------- | ------- | ----------------------------------------------- |
|
||||
| 2026-07-08 14:00 | classes | 实现分页查询,学到 Drizzle 的 limit/offset 用法 |
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 12. 安全与权限
|
||||
|
||||
### 12.1 分支保护规则(main)
|
||||
|
||||
- **禁止直接 push**:必须通过 PR
|
||||
- **必须 PR review**:至少 1 人(CODEOWNERS 自动分配)
|
||||
- **必须 CI 通过**:lint / typecheck / build
|
||||
- **禁止 force push**:开发 AI 无权,仅协调 AI 在事故时操作
|
||||
|
||||
### 12.2 AI 账号权限
|
||||
|
||||
| 角色 | push 特性分支 | 创建 PR | 合并 PR | push main | force push main |
|
||||
| ------- | ------------- | ------- | ----------- | --------- | --------------- |
|
||||
| 开发 AI | ✅ | ✅ | ❌ | ❌ | ❌ |
|
||||
| 协调 AI | ✅ | ✅ | ✅ | ❌ | ⚠️(仅事故) |
|
||||
| SRE AI | ✅(infra) | ✅ | ✅(infra) | ❌ | ⚠️(仅事故) |
|
||||
|
||||
### 12.3 敏感文件保护
|
||||
|
||||
以下文件修改需人类决策者额外审批:
|
||||
|
||||
- `.env`、`.env.example`(密钥相关)
|
||||
- `infra/security/secrets.example.env`
|
||||
- `infra/k8s/`(生产部署)
|
||||
- `.github/workflows/`(CI 配置)
|
||||
|
||||
---
|
||||
|
||||
## 13. 发布流程
|
||||
|
||||
### 13.1 版本号规则(见 git-workflow.md §7)
|
||||
|
||||
- 平台版本:`v{阶段}.{迭代}.{patch}`(如 `v1.2.0`)
|
||||
- 服务版本:`{service}:{semver}`(如 `classes:1.0.1`)
|
||||
|
||||
### 13.2 发布步骤(协调 AI 执行)
|
||||
|
||||
```bash
|
||||
# 1. 确认 main 稳定(CI 全绿)
|
||||
gh run list --branch main --limit 5
|
||||
|
||||
# 2. 打 tag
|
||||
git tag -a v1.2.0 -m "P1 阶段第 2 次迭代发布"
|
||||
git push origin v1.2.0
|
||||
|
||||
# 3. 触发 CI 构建镜像
|
||||
# (CI 由 tag push 触发,见 .github/workflows/)
|
||||
|
||||
# 4. 人类决策者确认部署到生产
|
||||
```
|
||||
|
||||
### 13.3 回滚(见 git-workflow.md §8)
|
||||
|
||||
```bash
|
||||
# K8s 回滚
|
||||
kubectl rollout undo deployment/api-gateway -n edu-system
|
||||
|
||||
# Git 回退(紧急)
|
||||
git revert <bad-commit>
|
||||
git push origin main
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 14. 快速参考卡
|
||||
|
||||
### 开发 AI 日常工作流
|
||||
|
||||
```bash
|
||||
# 拉新分支
|
||||
git checkout main && git pull && git checkout -b feat/<scope>-<task>-<ai-id>
|
||||
|
||||
# 开发 → 校验 → 提交
|
||||
pnpm run lint && pnpm run typecheck
|
||||
git add . && git commit -m "feat(<scope>): <subject>"
|
||||
|
||||
# 推送 → 创建 PR
|
||||
git push -u origin feat/<scope>-<task>-<ai-id>
|
||||
gh pr create --title "feat(<scope>): <subject>" --body "..."
|
||||
|
||||
# 等待审核 → 修改 → 重新推送
|
||||
# (协调 AI 合并后)
|
||||
git checkout main && git pull && git branch -d feat/<scope>-<task>-<ai-id>
|
||||
```
|
||||
|
||||
### 协调 AI 日常工作流
|
||||
|
||||
```bash
|
||||
# 查看待审核 PR
|
||||
gh pr list --state open --reviewer @edu-platform/arch
|
||||
|
||||
# 审核
|
||||
gh pr view <PR-NUMBER>
|
||||
gh pr checks <PR-NUMBER>
|
||||
|
||||
# 合并
|
||||
gh pr merge <PR-NUMBER> --squash --delete-branch
|
||||
|
||||
# 发布
|
||||
git tag -a v<version> -m "..." && git push origin v<version>
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 15. 相关文档
|
||||
|
||||
- [Git 工作流](./git-workflow.md) — 提交规范、分支策略、CODEOWNERS
|
||||
- [本地启动手册](./local-dev-runbook.md) — 手动启动服务
|
||||
- [项目规则](../../.trae/rules/project_rules.md) — 强制约束
|
||||
- [004 架构影响地图](../architecture/004_architecture_impact_map.md) — 模块与依赖关系
|
||||
- [CODEOWNERS](../../.github/CODEOWNERS) — Reviewer 自动分配
|
||||
- [PR 模板](../../.github/pull_request_template.md) — PR 必填项
|
||||
- [known-issues](../troubleshooting/known-issues.md) — 已知问题速查
|
||||
Reference in New Issue
Block a user