SQL agent with memory

工作流概述

这是一个包含13个节点的复杂工作流,主要用于自动化处理各种任务。

工作流源代码

下载
{
  "id": "AQJ6QnF2yVdCWMnx",
  "meta": {
    "instanceId": "fb924c73af8f703905bc09c9ee8076f48c17b596ed05b18c0ff86915ef8a7c4a",
    "templateCredsSetupCompleted": true
  },
  "name": "SQL agent with memory",
  "tags": [],
  "nodes": [
    {
      "id": "3544950e-4d8e-46ca-8f56-61c152a5cae3",
      "name": "Window Buffer Memory",
      "type": "@n8n/n8n-nodes-langchain.memoryBufferWindow",
      "position": [
        1220,
        500
      ],
      "parameters": {
        "contextWindowLength": 10
      },
      "typeVersion": 1.2
    },
    {
      "id": "743cc4e7-5f24-4adc-b872-7241ee775bd0",
      "name": "OpenAI Chat Model",
      "type": "@n8n/n8n-nodes-langchain.lmChatOpenAi",
      "position": [
        1000,
        500
      ],
      "parameters": {
        "model": "gpt-4-turbo",
        "options": {
          "temperature": 0.3
        }
      },
      "credentials": {
        "openAiApi": {
          "id": "rveqdSfp7pCRON1T",
          "name": "Ted's Tech Talks OpenAi"
        }
      },
      "typeVersion": 1
    },
    {
      "id": "cc30066c-ad2c-4729-82c1-a6b0f4214dee",
      "name": "When clicking \"Test workflow\"",
      "type": "n8n-nodes-base.manualTrigger",
      "position": [
        500,
        -80
      ],
      "parameters": {},
      "typeVersion": 1
    },
    {
      "id": "0deacd0d-45cb-4738-8da0-9d1251858867",
      "name": "Get chinook.zip example",
      "type": "n8n-nodes-base.httpRequest",
      "position": [
        700,
        -80
      ],
      "parameters": {
        "url": "https://www.sqlitetutorial.net/wp-content/uploads/2018/03/chinook.zip",
        "options": {}
      },
      "typeVersion": 4.2
    },
    {
      "id": "61f34708-f8ed-44a9-8522-6042d28511ae",
      "name": "Extract zip file",
      "type": "n8n-nodes-base.compression",
      "position": [
        900,
        -80
      ],
      "parameters": {},
      "typeVersion": 1.1
    },
    {
      "id": "6a12d9ac-f1b7-4267-8b34-58cdb9d347bb",
      "name": "Save chinook.db locally",
      "type": "n8n-nodes-base.readWriteFile",
      "position": [
        1100,
        -80
      ],
      "parameters": {
        "options": {},
        "fileName": "./chinook.db",
        "operation": "write",
        "dataPropertyName": "file_0"
      },
      "typeVersion": 1
    },
    {
      "id": "701d1325-4186-4185-886a-3738163db603",
      "name": "Load local chinook.db",
      "type": "n8n-nodes-base.readWriteFile",
      "position": [
        620,
        360
      ],
      "parameters": {
        "options": {},
        "fileSelector": "./chinook.db"
      },
      "typeVersion": 1
    },
    {
      "id": "d7b3813d-8180-4ff1-87a4-bd54a03043af",
      "name": "Sticky Note",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        440,
        -280.9454545454546
      ],
      "parameters": {
        "width": 834.3272727272731,
        "height": 372.9454545454546,
        "content": "## Run this part only once
This section:
* downloads the example zip file from https://www.sqlitetutorial.net/sqlite-sample-database/
* extracts the archive (it contains only a single file)
* saves the extracted `chinook.db` SQLite database locally

Now you can use chat to \"talk\" to your data!"
      },
      "typeVersion": 1
    },
    {
      "id": "6bd25563-2c59-44c2-acf9-407bd28a15cf",
      "name": "Sticky Note1",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        400,
        240
      ],
      "parameters": {
        "width": 558.5454545454544,
        "height": 297.89090909090913,
        "content": "## On every chat message:
* the local SQLite database is loaded
* JSON from Chat Trigger is combined with SQLite binary data"
      },
      "typeVersion": 1
    },
    {
      "id": "2be63956-236e-46f7-b8e4-0f55e2e25a5c",
      "name": "Combine chat input with the binary",
      "type": "n8n-nodes-base.set",
      "position": [
        820,
        360
      ],
      "parameters": {
        "mode": "raw",
        "options": {
          "includeBinary": true
        },
        "jsonOutput": "={{ $('Chat Trigger').item.json }}
"
      },
      "typeVersion": 3.3
    },
    {
      "id": "7f4c9adb-eab4-40d7-ad2e-44f2c0e3e30a",
      "name": "Sticky Note2",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        980,
        120
      ],
      "parameters": {
        "width": 471.99692219161466,
        "height": 511.16641410437836,
        "content": "### LangChain SQL Agent can make several queries before producing the final answer.
Try these examples:
1. \"Please describe the database\". This input usually requires just 1 query + an extra observation to produce a final answer.
2. \"What are the revenues by genre?\". This input will launch a series of Agent actions, because it needs to make several queries.

The final answer is stored in the memory and will be recalled on the next input from the user."
      },
      "typeVersion": 1
    },
    {
      "id": "ac819eb5-13b2-4280-b9d6-06ec1209700e",
      "name": "AI Agent",
      "type": "@n8n/n8n-nodes-langchain.agent",
      "position": [
        1020,
        360
      ],
      "parameters": {
        "agent": "sqlAgent",
        "options": {},
        "dataSource": "sqlite"
      },
      "typeVersion": 1.6
    },
    {
      "id": "5ecaa3eb-e93e-4e41-bbc0-98a8c2b2d463",
      "name": "Chat Trigger",
      "type": "@n8n/n8n-nodes-langchain.chatTrigger",
      "position": [
        420,
        360
      ],
      "webhookId": "fb565f08-a459-4ff9-8249-1ede58599660",
      "parameters": {},
      "typeVersion": 1
    }
  ],
  "active": false,
  "pinData": {},
  "settings": {
    "executionOrder": "v1"
  },
  "versionId": "fbc06ddd-dbd8-49ee-bbee-2f495d5651a2",
  "connections": {
    "Chat Trigger": {
      "main": [
        [
          {
            "node": "Load local chinook.db",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Extract zip file": {
      "main": [
        [
          {
            "node": "Save chinook.db locally",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "OpenAI Chat Model": {
      "ai_languageModel": [
        [
          {
            "node": "AI Agent",
            "type": "ai_languageModel",
            "index": 0
          }
        ]
      ]
    },
    "Window Buffer Memory": {
      "ai_memory": [
        [
          {
            "node": "AI Agent",
            "type": "ai_memory",
            "index": 0
          }
        ]
      ]
    },
    "Load local chinook.db": {
      "main": [
        [
          {
            "node": "Combine chat input with the binary",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Get chinook.zip example": {
      "main": [
        [
          {
            "node": "Extract zip file",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "When clicking \"Test workflow\"": {
      "main": [
        [
          {
            "node": "Get chinook.zip example",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Combine chat input with the binary": {
      "main": [
        [
          {
            "node": "AI Agent",
            "type": "main",
            "index": 0
          }
        ]
      ]
    }
  }
}

功能特点

  • 自动检测新邮件
  • AI智能内容分析
  • 自定义分类规则
  • 批量处理能力
  • 详细的处理日志

技术分析

节点类型及作用

  • @N8N/N8N Nodes Langchain.Memorybufferwindow
  • @N8N/N8N Nodes Langchain.Lmchatopenai
  • Manualtrigger
  • Httprequest
  • Compression

复杂度评估

配置难度:
★★★★☆
维护难度:
★★☆☆☆
扩展性:
★★★★☆

实施指南

前置条件

  • 有效的Gmail账户
  • n8n平台访问权限
  • Google API凭证
  • AI分类服务订阅

配置步骤

  1. 在n8n中导入工作流JSON文件
  2. 配置Gmail节点的认证信息
  3. 设置AI分类器的API密钥
  4. 自定义分类规则和标签映射
  5. 测试工作流执行
  6. 配置定时触发器(可选)

关键参数

参数名称 默认值 说明
maxEmails 50 单次处理的最大邮件数量
confidenceThreshold 0.8 分类置信度阈值
autoLabel true 是否自动添加标签

最佳实践

优化建议

  • 定期更新AI分类模型以提高准确性
  • 根据邮件量调整处理批次大小
  • 设置合理的分类置信度阈值
  • 定期清理过期的分类规则

安全注意事项

  • 妥善保管API密钥和认证信息
  • 限制工作流的访问权限
  • 定期审查处理日志
  • 启用双因素认证保护Gmail账户

性能优化

  • 使用增量处理减少重复工作
  • 缓存频繁访问的数据
  • 并行处理多个邮件分类任务
  • 监控系统资源使用情况

故障排除

常见问题

邮件未被正确分类

检查AI分类器的置信度阈值设置,适当降低阈值或更新训练数据。

Gmail认证失败

确认Google API凭证有效且具有正确的权限范围,重新进行OAuth授权。

调试技巧

  • 启用详细日志记录查看每个步骤的执行情况
  • 使用测试邮件验证分类逻辑
  • 检查网络连接和API服务状态
  • 逐步执行工作流定位问题节点

错误处理

工作流包含以下错误处理机制:

  • 网络超时自动重试(最多3次)
  • API错误记录和告警
  • 处理失败邮件的隔离机制
  • 异常情况下的回滚操作