LibreFang

工作流

LibreFang 工作流系统允许您将多个 Agent 连接在一起，创建复杂的自动化流程。

概述

工作流是由触发器、条件、动作和 Agent 组成的可配置流程。

触发器 → 条件检查 → Agent 执行 → 条件检查 → Agent 执行 → ... → 结束

工作流结构

{
  "name": "lead-generation",
  "version": "1.0.0",
  "trigger": {
    "type": "schedule",
    "cron": "0 9 * * *"
  },
  "steps": [
    {
      "id": "research",
      "agent": "researcher",
      "input": "Find companies matching ICP: {{icp}}"
    },
    {
      "id": "enrich",
      "agent": "analyst",
      "input": "Enrich leads with {{research.results}}"
    },
    {
      "id": "score",
      "agent": "classifier",
      "input": "Score leads: {{enrich.results}}"
    }
  ],
  "on_complete": {
    "action": "notify",
    "channel": "telegram",
    "message": "Generated {{score.count}} qualified leads"
  }
}

触发器类型

1. 定时触发

{
  "type": "schedule",
  "cron": "0 9 * * *"
}

Cron 表达式：

表达式说明
0 9 * * *每天 9:00
0 9 * * 1-5工作日 9:00
*/15 * * * *每 15 分钟
0 0 1 * *每月 1 日

2. 事件触发

{
  "type": "event",
  "event": "email_received",
  "filter": {
    "from": "*.@example.com"
  }
}

3. Webhook 触发

{
  "type": "webhook",
  "path": "/webhook/github",
  "method": "POST"
}

4. 手动触发

{
  "type": "manual"
}

步骤类型

Agent 步骤

{
  "id": "step1",
  "type": "agent",
  "agent": "researcher",
  "input": "Research {{context.query}}",
  "output_var": "research_results"
}

条件步骤

{
  "id": "check",
  "type": "condition",
  "expression": "{{research_results.count}} > 10"
}

延迟步骤

{
  "id": "wait",
  "type": "delay",
  "duration": "1h"
}

通知步骤

{
  "id": "notify",
  "type": "notification",
  "channel": "telegram",
  "message": "Workflow complete!"
}

条件逻辑

IF/ELSE

{
  "type": "condition",
  "if": "{{count}} > 10",
  "then": {
    "agent": "big_team"
  },
  "else": {
    "agent": "small_team"
  }
}

SWITCH

{
  "type": "switch",
  "value": "{{priority}}",
  "cases": {
    "high": { "agent": "urgent_team" },
    "medium": { "agent": "normal_team" },
    "low": { "agent": "background_team" }
  }
}

变量

输入变量

{
  "input": "Hello {{name}}"
}

输出变量

{
  "output_var": "step1_result"
}

全局变量

{
  "variables": {
    "icp": "companies in AI space, >100 employees",
    "target_count": 50
  }
}

错误处理

重试

{
  "id": "step1",
  "retry": {
    "max_attempts": 3,
    "delay": "5s",
    "backoff": "exponential"
  }
}

错误处理

{
  "on_error": {
    "fallback_agent": "fallback",
    "notify": true
  }
}

CLI 命令

# 列出工作流
librefang workflow list

# 创建工作流
librefang workflow create workflow.json

# 运行工作流
librefang workflow run <id> <input>

# 查看工作流状态
librefang workflow status <id>

# 删除工作流
librefang workflow delete <id>

示例工作流

每日线索生成

{
  "name": "daily-lead-generation",
  "trigger": {
    "type": "schedule",
    "cron": "0 9 * * *"
  },
  "steps": [
    {
      "id": "discover",
      "agent": "lead_discovery",
      "input": "Find companies matching our ICP"
    },
    {
      "id": "enrich",
      "agent": "data_enrichment",
      "input": "Enrich with {{discover.companies}}"
    },
    {
      "id": "score",
      "agent": "lead_scoring",
      "input": "Score: {{enrich.leads}}"
    }
  ],
  "on_complete": {
    "action": "notify",
    "channel": "telegram",
    "message": "Generated {{score.count}} leads"
  }
}

客户支持路由

{
  "name": "support-routing",
  "trigger": {
    "type": "webhook",
    "path": "/webhook/support"
  },
  "steps": [
    {
      "id": "classify",
      "agent": "ticket_classifier",
      "input": "Classify: {{input.subject}}"
    },
    {
      "id": "route",
      "type": "switch",
      "value": "{{classify.category}}",
      "cases": {
        "billing": { "agent": "billing_team" },
        "technical": { "agent": "tech_team" },
        "general": { "agent": "general_team" }
      }
    }
  ]
}

最佳实践

  1. 保持简单 - 避免过于复杂的工作流
  2. 添加重试 - 网络可能不稳定
  3. 设置超时 - 防止步骤卡住
  4. 记录日志 - 便于调试
  5. 监控执行 - 跟踪工作流状态