327 lines
6.2 KiB
Markdown
327 lines
6.2 KiB
Markdown
# AI 工作流配置
|
||
|
||
> 使用 OpenClaw 工作流引擎,构建定时任务与自动化业务流。
|
||
|
||
## 🔄 什么是工作流?
|
||
|
||
工作流(Workflow)是预定义的一系列自动化步骤,可以:
|
||
|
||
- ⏰ **定时执行**:每天早上发送报表
|
||
- 🔗 **事件触发**:用户注册后自动发送欢迎邮件
|
||
- 🔁 **条件分支**:根据条件执行不同操作
|
||
- 📊 **数据处理**:批量处理数据
|
||
|
||
## 🏗️ 核心概念
|
||
|
||
| 概念 | 说明 |
|
||
|------|------|
|
||
| **触发器 (Trigger)** | 工作流的启动条件 |
|
||
| **节点 (Node)** | 工作流中的单个步骤 |
|
||
| **连接 (Edge)** | 节点之间的数据流向 |
|
||
| **变量 (Variable)** | 存储和传递数据 |
|
||
| **执行日志** | 每次运行的记录 |
|
||
|
||
## 🚀 创建第一个工作流
|
||
|
||
### 方式一:代码定义
|
||
|
||
```typescript
|
||
import { WorkflowBuilder } from '@websopy/ai-sdk'
|
||
|
||
const workflow = new WorkflowBuilder({
|
||
name: '每日数据报告',
|
||
description: '每天早上 9 点生成数据报告'
|
||
})
|
||
|
||
// 添加触发器:定时执行
|
||
workflow.addTrigger({
|
||
type: 'schedule',
|
||
config: {
|
||
cron: '0 9 * * *', // 每天 9:00
|
||
timezone: 'Asia/Shanghai'
|
||
}
|
||
})
|
||
|
||
// 添加节点:获取数据
|
||
workflow.addNode('fetch-data', {
|
||
type: 'http-request',
|
||
config: {
|
||
url: 'https://api.analytics.com/daily-stats',
|
||
method: 'GET'
|
||
}
|
||
})
|
||
|
||
// 添加节点:生成报告
|
||
workflow.addNode('generate-report', {
|
||
type: 'ai-agent',
|
||
config: {
|
||
prompt: '根据以下数据生成日报:{{fetch-data.output}}'
|
||
}
|
||
})
|
||
|
||
// 添加节点:发送邮件
|
||
workflow.addNode('send-email', {
|
||
type: 'email',
|
||
config: {
|
||
to: ['team@company.com'],
|
||
subject: '📊 每日数据报告',
|
||
body: '{{generate-report.output}}'
|
||
}
|
||
})
|
||
|
||
// 节点连接
|
||
workflow.connect('fetch-data', 'generate-report')
|
||
workflow.connect('generate-report', 'send-email')
|
||
|
||
// 保存
|
||
const created = await client.workflow.create(workflow.toJSON())
|
||
console.log('工作流 ID:', created.id)
|
||
```
|
||
|
||
### 方式二:可视化编辑器
|
||
|
||
在控制台 → AI 功能 → 工作流,点击「新建工作流」打开可视化编辑器。
|
||
|
||
## 📦 节点类型
|
||
|
||
### 1. HTTP 请求
|
||
|
||
```typescript
|
||
workflow.addNode('call-api', {
|
||
type: 'http-request',
|
||
config: {
|
||
url: 'https://api.example.com/data',
|
||
method: 'POST',
|
||
headers: {
|
||
'Authorization': 'Bearer xxx'
|
||
},
|
||
body: {
|
||
query: 'SELECT * FROM orders'
|
||
}
|
||
}
|
||
})
|
||
```
|
||
|
||
### 2. AI 智能体
|
||
|
||
```typescript
|
||
workflow.addNode('analyze', {
|
||
type: 'ai-agent',
|
||
config: {
|
||
model: 'gpt-4',
|
||
prompt: '分析以下数据,找出异常:{{input}}',
|
||
temperature: 0.5
|
||
}
|
||
})
|
||
```
|
||
|
||
### 3. 条件分支
|
||
|
||
```typescript
|
||
workflow.addNode('check-status', {
|
||
type: 'condition',
|
||
config: {
|
||
conditions: [
|
||
{
|
||
expression: '{{input.amount}} > 10000',
|
||
nextNode: 'high-value-handler'
|
||
},
|
||
{
|
||
expression: '{{input.amount}} > 1000',
|
||
nextNode: 'medium-handler'
|
||
}
|
||
],
|
||
default: 'low-value-handler'
|
||
}
|
||
})
|
||
```
|
||
|
||
### 4. 循环
|
||
|
||
```typescript
|
||
workflow.addNode('process-items', {
|
||
type: 'loop',
|
||
config: {
|
||
items: '{{batch-data.items}}',
|
||
maxIterations: 100,
|
||
// 循环体节点
|
||
nodes: [...]
|
||
}
|
||
})
|
||
```
|
||
|
||
### 5. 等待/延迟
|
||
|
||
```typescript
|
||
workflow.addNode('delay', {
|
||
type: 'delay',
|
||
config: {
|
||
duration: '24h' // 或 3600000(毫秒)
|
||
}
|
||
})
|
||
```
|
||
|
||
### 6. Webhook
|
||
|
||
```typescript
|
||
workflow.addNode('notify', {
|
||
type: 'webhook',
|
||
config: {
|
||
url: 'https://your-server.com/webhook',
|
||
method: 'POST',
|
||
headers: {
|
||
'Content-Type': 'application/json'
|
||
},
|
||
body: {
|
||
event: 'workflow.completed',
|
||
data: '{{workflow-output}}'
|
||
}
|
||
}
|
||
})
|
||
```
|
||
|
||
## 🔔 触发器类型
|
||
|
||
### 定时触发
|
||
|
||
```typescript
|
||
workflow.addTrigger({
|
||
type: 'schedule',
|
||
config: {
|
||
cron: '0 */4 * * *', // 每 4 小时
|
||
// 或使用简单表达式
|
||
interval: 'daily', // daily, weekly, monthly
|
||
time: '09:00'
|
||
}
|
||
})
|
||
```
|
||
|
||
### Webhook 触发
|
||
|
||
```typescript
|
||
workflow.addTrigger({
|
||
type: 'webhook',
|
||
config: {
|
||
// 创建后获得 webhook URL
|
||
}
|
||
})
|
||
```
|
||
|
||
### 事件触发
|
||
|
||
```typescript
|
||
workflow.addTrigger({
|
||
type: 'event',
|
||
config: {
|
||
events: [
|
||
'user.created',
|
||
'order.completed',
|
||
'ai.task.failed'
|
||
]
|
||
}
|
||
})
|
||
```
|
||
|
||
## 🔧 变量与数据处理
|
||
|
||
### 定义变量
|
||
|
||
```typescript
|
||
workflow.addNode('process', {
|
||
type: 'transform',
|
||
config: {
|
||
variables: {
|
||
totalAmount: '{{input.orders}} | sum("amount")',
|
||
avgScore: '{{input.reviews}} | avg("score")',
|
||
userNames: '{{input.users}} | map("name")'
|
||
}
|
||
}
|
||
})
|
||
```
|
||
|
||
### 数据转换
|
||
|
||
```typescript
|
||
workflow.addNode('transform', {
|
||
type: 'transform',
|
||
config: {
|
||
// 过滤器
|
||
filter: '{{data}} | filter(item => item.active == true)',
|
||
// 映射
|
||
map: '{{data}} | map(item => ({ id: item.id, name: item.name }))',
|
||
// 排序
|
||
sort: '{{data}} | sortBy("created_at", "desc")'
|
||
}
|
||
})
|
||
```
|
||
|
||
## 📊 执行与监控
|
||
|
||
### 手动执行
|
||
|
||
```typescript
|
||
// 立即执行一次
|
||
const run = await client.workflow.run(workflowId, {
|
||
input: {
|
||
date: '2024-01-15'
|
||
}
|
||
})
|
||
|
||
console.log('执行 ID:', run.id)
|
||
```
|
||
|
||
### 查看执行日志
|
||
|
||
```typescript
|
||
const logs = await client.workflow.getRunLogs(run.id)
|
||
|
||
for (const log of logs) {
|
||
console.log(`[${log.timestamp}] ${log.node}: ${log.status}`)
|
||
if (log.error) {
|
||
console.log(' 错误:', log.error)
|
||
}
|
||
}
|
||
```
|
||
|
||
### 执行状态
|
||
|
||
```typescript
|
||
const status = await client.workflow.getRunStatus(run.id)
|
||
|
||
console.log('状态:', status.status) // running, completed, failed
|
||
console.log('当前节点:', status.currentNode)
|
||
console.log('已用时间:', status.elapsedTime)
|
||
```
|
||
|
||
## ❓ 常见问题
|
||
|
||
### Q: 工作流执行失败怎么办?
|
||
|
||
1. 查看执行日志定位问题节点
|
||
2. 检查输入数据格式
|
||
3. 查看节点配置是否正确
|
||
4. 联系支持时提供执行 ID
|
||
|
||
### Q: 如何调试工作流?
|
||
|
||
使用「测试执行」功能,传入测试数据,逐步查看每个节点的输出。
|
||
|
||
### Q: 可以嵌套工作流吗?
|
||
|
||
是的,可以在一个工作流中调用另一个工作流:
|
||
|
||
```typescript
|
||
workflow.addNode('sub-workflow', {
|
||
type: 'workflow',
|
||
config: {
|
||
workflowId: 'wf-xxx',
|
||
input: '{{current-data}}'
|
||
}
|
||
})
|
||
```
|
||
|
||
---
|
||
|
||
**上一步:** [RAG 知识库搭建](./rag.md)
|
||
**下一步:** [Docker Compose 部署](./docker.md)
|