OpenAI-model-examples
工作流概述
这是一个包含27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
{
"id": "147",
"meta": {
"instanceId": "dfdeafd1c3ed2ee08eeab8c2fa0c3f522066931ed8138ccd35dc20a1e69decd3"
},
"name": "OpenAI-model-examples",
"tags": [],
"nodes": [
{
"id": "ad6dc2cd-21cc-4563-86ba-f78cc4a55543",
"name": "When clicking \"Execute Workflow\"",
"type": "n8n-nodes-base.manualTrigger",
"position": [
-640,
380
],
"parameters": {},
"typeVersion": 1
},
{
"id": "b370da23-ead4-4221-b7fe-a9d943f7fbb9",
"name": "davinci-003-complete",
"type": "n8n-nodes-base.openAi",
"position": [
1160,
60
],
"parameters": {
"prompt": "={{ $json.text }}
Tl;dr:",
"options": {
"maxTokens": 500
}
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 1
},
{
"id": "5e04f355-36c0-4540-8e65-68118cb73135",
"name": "ChatGPT-ex2",
"type": "n8n-nodes-base.openAi",
"position": [
1160,
740
],
"parameters": {
"prompt": {
"messages": [
{
"role": "system",
"content": "=You are an assistant. Always add 5 emojis to the end of your answer."
},
{
"content": "=Write tl;dr of the wollowing text: {{ $json.text}}"
}
]
},
"options": {
"maxTokens": 500,
"temperature": 0.8
},
"resource": "chat"
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 1
},
{
"id": "16a7cf80-16e3-44f9-b15c-7501417fe38f",
"name": "davinci-003-edit",
"type": "n8n-nodes-base.openAi",
"position": [
1340,
60
],
"parameters": {
"input": "={{ $json.text }}",
"options": {},
"operation": "edit",
"instruction": "translate to German"
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 1
},
{
"id": "95254870-65c3-4714-83fb-20ba2c0ca007",
"name": "ChatGPT-ex1.1",
"type": "n8n-nodes-base.openAi",
"position": [
1160,
380
],
"parameters": {
"prompt": {
"messages": [
{
"content": "=Write a Tl;dr of the followint text: {{ $json.text }}"
}
]
},
"options": {
"maxTokens": 500
},
"resource": "chat"
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 1
},
{
"id": "be9c4820-18b0-46fd-a5a0-51a5dc3ebed5",
"name": "ChatGPT-ex1.2",
"type": "n8n-nodes-base.openAi",
"position": [
1340,
380
],
"parameters": {
"prompt": {
"messages": [
{
"content": "=Translate to German the following text: {{ $json.message.content }}"
}
]
},
"options": {
"maxTokens": 500
},
"resource": "chat"
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 1
},
{
"id": "c52c875b-5270-44ac-bfca-ce25124e3d04",
"name": "Text-example",
"type": "n8n-nodes-base.code",
"position": [
540,
380
],
"parameters": {
"jsCode": "return [
{
\"text\": \"Science Underground with your host, Anissa Ramirez. In this episode, how to stop your bathroom mirror from fogging up with a little dash of science. I'm Anissa Ramirez and this is Science Underground. We've all been there. You come out of the shower and you go to the mirror and you can't see yourself because the mirror is fogged up. You can't see anything until you first clear off the surface. Every morning it's the same thing. Shower, fog, shower, fog, shower, fog. There's gotta be a better way. Well, there is. Before you take the next shower, wipe a bit of shaving cream on the surface of the mirror and keep it there for about 30 seconds. Then wipe it off. The next time you take a shower, that part of the mirror that was covered with shaving cream will be amazingly fog free. And the shaving cream will keep the water from fogging up for a few weeks. So what's going on? Well, the fog on your mirror is made out of little itty bitty water droplets. If you were to look at the surface of the mirror under the microscope, you will see that the surface looks like a newly waxed car. The water forms beads, preventing you from seeing yourself in the mirror. When you add shaving cream to the surface of the mirror, the water droplets are no longer beads. They are a thin, smoothed out layer of water. Just like the surface of an old car that hasn't been waxed. Scientists would say that the shaving cream has changed the surface tension of the mirror. So there you have it. There's the answer. The secret to fogless mirrors is shaving cream. A little dab of science will do you. I'm Anissa Ramirez, and this was Science Underground.\"
}
];"
},
"typeVersion": 1
},
{
"id": "45d3bad7-0e9a-426b-b4e9-b3568181d9dc",
"name": "Code-ex3.1",
"type": "n8n-nodes-base.code",
"position": [
1160,
1100
],
"parameters": {
"jsCode": "var intext = $input.first().json;
var messages = [
{\"role\": \"system\", \"content\": \"You are a helpful assistant. Write a Tl;dr of each user message\"},
{\"role\": \"user\", \"content\": intext.text}
];
return {\"messages\":messages};"
},
"typeVersion": 1
},
{
"id": "4db3de05-51a7-46ea-a818-508bdcb04582",
"name": "ChatGPT-ex3.1",
"type": "n8n-nodes-base.httpRequest",
"position": [
1340,
1100
],
"parameters": {
"url": "https://api.openai.com/v1/chat/completions",
"method": "POST",
"options": {},
"sendBody": true,
"authentication": "predefinedCredentialType",
"bodyParameters": {
"parameters": [
{
"name": "model",
"value": "gpt-3.5-turbo"
},
{
"name": "temperature",
"value": "={{ parseFloat(0.8) }}"
},
{
"name": "n",
"value": "={{ Number(1) }}"
},
{
"name": "max_tokens",
"value": "={{ Number(500) }}"
},
{
"name": "messages",
"value": "={{ $json.messages }}"
}
]
},
"nodeCredentialType": "openAiApi"
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 3
},
{
"id": "709fcd7c-deb3-469d-b16b-62d4d36d100d",
"name": "ChatGPT-ex3.2",
"type": "n8n-nodes-base.openAi",
"position": [
1880,
1100
],
"parameters": {
"prompt": {
"messages": [
{
"role": "system",
"content": "=You are now a DALLE-2 prompt generation tool that will generate a suitable prompt. Write a promt to create a cover image relevant to the user input. The image should be in a comic style of the 60-s."
},
{
"content": "={{ $json.choices[0].message.content }}"
}
]
},
"options": {
"maxTokens": 500,
"temperature": 0.8
},
"resource": "chat"
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 1
},
{
"id": "6b32cc45-5ba2-4605-b690-3929ec9acecf",
"name": "Sticky Note",
"type": "n8n-nodes-base.stickyNote",
"position": [
900,
-60
],
"parameters": {
"width": 746.6347949130579,
"height": 295.50954755505853,
"content": "## The old way of using text completion and text edit
### Davinci model is 10 times more expensive then ChatGPT, consider switching to the new API:
https://openai.com/blog/introducing-chatgpt-and-whisper-apis
"
},
"typeVersion": 1
},
{
"id": "3cc74d77-7b02-40fd-83d8-f540d5ff34ab",
"name": "Sticky Note1",
"type": "n8n-nodes-base.stickyNote",
"position": [
-160,
260
],
"parameters": {
"width": 428.4578974150008,
"height": 316.6202633391793,
"content": "## Whisper-1 example
### Prepare your audio file and send it to whisper-1 transcription model"
},
"typeVersion": 1
},
{
"id": "6ba8069a-485c-497c-8b27-4c7562fbccab",
"name": "Sticky Note2",
"type": "n8n-nodes-base.stickyNote",
"position": [
380,
280
],
"parameters": {
"width": 421.9002034748082,
"height": 302.4086532331564,
"content": "## An example of transcribed text
### Please pause this node when using real audio files"
},
"typeVersion": 1
},
{
"id": "c71001e6-b80f-41dd-bcdd-10927014b374",
"name": "Sticky Note3",
"type": "n8n-nodes-base.stickyNote",
"position": [
900,
280
],
"parameters": {
"width": 747.8556016477869,
"height": 288.18470714667706,
"content": "## ChatGPT example 1.1 and 1.2
### Write a Tl;dr of the text input
### Translate it to German
### only user content provided"
},
"typeVersion": 1
},
{
"id": "4605be68-4c57-404f-8624-e095c8e86ff9",
"name": "Sticky Note4",
"type": "n8n-nodes-base.stickyNote",
"position": [
900,
620
],
"parameters": {
"width": 742.9723747088658,
"height": 288.18470714667706,
"content": "## ChatGPT example 2
### Use system content to provide general instruction
### Manual setup of system and user content"
},
"typeVersion": 1
},
{
"id": "f5b72d7a-655a-4cc9-b722-b75429889d1d",
"name": "Sticky Note5",
"type": "n8n-nodes-base.stickyNote",
"position": [
900,
960
],
"parameters": {
"width": 739.309954504675,
"height": 288.18470714667706,
"content": "## ChatGPT example 3.1
### When using ChatGPT programmatically, create an array of system / user / assistant contents and append them one after another
### Call ChatGPT API via HTTP Request node to provide all messages at once"
},
"typeVersion": 1
},
{
"id": "a003a4db-1960-4867-8dfe-3114cf0742f3",
"name": "DALLE-ex3.3",
"type": "n8n-nodes-base.openAi",
"position": [
2060,
1100
],
"parameters": {
"prompt": "={{ $json.message.content }}",
"options": {
"n": 4,
"size": "512x512"
},
"resource": "image"
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 1
},
{
"id": "d71a01ff-4d47-4675-964c-c47820d3989b",
"name": "Sticky Note6",
"type": "n8n-nodes-base.stickyNote",
"position": [
1720,
960
],
"parameters": {
"width": 611.1252473579985,
"height": 284.52228694248623,
"content": "## ChatGPT example 3.2 & DALLE-2 example 3.3
### Use ChatGPT to create a prompt for a cover image of the Tl;dr message
### Use OpenAI node to generate 4 images using the auto-generated prompt"
},
"typeVersion": 1
},
{
"id": "f5a55cfe-c110-4833-9668-1f1ba895860f",
"name": "ChatGPT-ex4",
"type": "n8n-nodes-base.openAi",
"position": [
1240,
1420
],
"parameters": {
"model": "gpt-3.5-turbo-0301",
"prompt": {
"messages": [
{
"content": "={{ $json.prompt }}"
}
]
},
"options": {
"maxTokens": 500,
"temperature": 0.5
},
"resource": "chat"
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 1
},
{
"id": "8a9f7a20-187c-4494-8005-b10d066d04e2",
"name": "Set-ex4",
"type": "n8n-nodes-base.set",
"position": [
1060,
1420
],
"parameters": {
"values": {
"string": [
{
"name": "model",
"value": "code-davinci-002"
},
{
"name": "suffix",
"value": "</svg>"
},
{
"name": "prompt",
"value": "=Create an HTML code with and SVG tag that contains random shapes of various colors. Include triangles, lines, ellipses and other shapes"
}
]
},
"options": {},
"keepOnlySet": true
},
"typeVersion": 1
},
{
"id": "68fcc6a2-761c-42ac-8778-313c8db7d53c",
"name": "HTML-ex4",
"type": "n8n-nodes-base.html",
"position": [
1420,
1420
],
"parameters": {
"html": "{{$json.message.content }}"
},
"typeVersion": 1
},
{
"id": "1f70cf3f-b6a9-4ea7-9486-c7565e6951b7",
"name": "Sticky Note7",
"type": "n8n-nodes-base.stickyNote",
"position": [
900,
1300
],
"parameters": {
"width": 739.309954504675,
"height": 288.18470714667706,
"content": "## ChatGPT example 4
### Generate HTML code that contains SVG image"
},
"typeVersion": 1
},
{
"id": "d857acd9-ea74-44d2-ac89-66b1fac4645f",
"name": "Sticky Note8",
"type": "n8n-nodes-base.stickyNote",
"position": [
900,
1640
],
"parameters": {
"width": 739.309954504675,
"height": 288.18470714667706,
"content": "## ChatGPT example 5
### Provide several outputs. Useful for quick replies (i.e. in Gmail / Outlook)"
},
"typeVersion": 1
},
{
"id": "fe64533a-4cd4-4adc-a48a-8abf3f2d61d7",
"name": "ChatGPT-ex",
"type": "n8n-nodes-base.openAi",
"position": [
1160,
1760
],
"parameters": {
"model": "gpt-3.5-turbo-0301",
"prompt": {
"messages": [
{
"role": "system",
"content": "Act as an e-mail client. Provide a five to eight word answers to a given user messages."
},
{
"content": "Hi There! My name is Jack.
I'm sending you an overview of my pricelist attached.
Could you please reply to me within 3 days?
Best regards and have a nice day,
Jack"
}
]
},
"options": {
"n": 3,
"maxTokens": 15,
"temperature": 0.8
},
"resource": "chat"
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 1
},
{
"id": "6c9f8a70-99ae-4310-8e6a-26cc6f75b3a2",
"name": "LoadMP3",
"type": "n8n-nodes-base.readBinaryFiles",
"disabled": true,
"position": [
-80,
380
],
"parameters": {
"fileSelector": "/home/node/.n8n/OpenAI-article/Using Science to Stop Your Mirror From Fogging Up.mp3"
},
"typeVersion": 1
},
{
"id": "0edc1996-6484-4e62-a47b-5666dfbb3546",
"name": "Whisper-transcribe",
"type": "n8n-nodes-base.httpRequest",
"disabled": true,
"position": [
100,
380
],
"parameters": {
"url": "https://api.openai.com/v1/audio/transcriptions",
"method": "POST",
"options": {},
"sendBody": true,
"contentType": "multipart-form-data",
"authentication": "predefinedCredentialType",
"bodyParameters": {
"parameters": [
{
"name": "model",
"value": "whisper-1"
},
{
"name": "file",
"parameterType": "formBinaryData",
"inputDataFieldName": "data"
}
]
},
"nodeCredentialType": "openAiApi"
},
"credentials": {
"openAiApi": {
"id": "63",
"name": "OpenAi account"
}
},
"typeVersion": 3
},
{
"id": "c12ba294-bdcd-4ece-8370-fa6a83a8ef0b",
"name": "Sticky Note9",
"type": "n8n-nodes-base.stickyNote",
"position": [
-840,
260
],
"parameters": {
"width": 596.9600747621192,
"height": 320.63203364295396,
"content": "## Do not run the whole workflow, it's rather slow
### Better execute the last node of each branch or simply disconnect branches that are not needed"
},
"typeVersion": 1
}
],
"active": false,
"pinData": {},
"settings": {
"callerPolicy": "workflowsFromSameOwner",
"saveManualExecutions": false,
"saveDataSuccessExecution": "none"
},
"versionId": "972cd971-9e7e-4a1d-b3fb-6f061e23e96f",
"connections": {
"LoadMP3": {
"main": [
[
{
"node": "Whisper-transcribe",
"type": "main",
"index": 0
}
]
]
},
"Set-ex4": {
"main": [
[
{
"node": "ChatGPT-ex4",
"type": "main",
"index": 0
}
]
]
},
"Code-ex3.1": {
"main": [
[
{
"node": "ChatGPT-ex3.1",
"type": "main",
"index": 0
}
]
]
},
"ChatGPT-ex4": {
"main": [
[
{
"node": "HTML-ex4",
"type": "main",
"index": 0
}
]
]
},
"Text-example": {
"main": [
[
{
"node": "davinci-003-complete",
"type": "main",
"index": 0
},
{
"node": "ChatGPT-ex1.1",
"type": "main",
"index": 0
},
{
"node": "ChatGPT-ex2",
"type": "main",
"index": 0
},
{
"node": "Code-ex3.1",
"type": "main",
"index": 0
}
]
]
},
"ChatGPT-ex1.1": {
"main": [
[
{
"node": "ChatGPT-ex1.2",
"type": "main",
"index": 0
}
]
]
},
"ChatGPT-ex3.1": {
"main": [
[
{
"node": "ChatGPT-ex3.2",
"type": "main",
"index": 0
}
]
]
},
"ChatGPT-ex3.2": {
"main": [
[
{
"node": "DALLE-ex3.3",
"type": "main",
"index": 0
}
]
]
},
"Whisper-transcribe": {
"main": [
[
{
"node": "Text-example",
"type": "main",
"index": 0
}
]
]
},
"davinci-003-complete": {
"main": [
[
{
"node": "davinci-003-edit",
"type": "main",
"index": 0
}
]
]
},
"When clicking \"Execute Workflow\"": {
"main": [
[
{
"node": "LoadMP3",
"type": "main",
"index": 0
},
{
"node": "Set-ex4",
"type": "main",
"index": 0
},
{
"node": "ChatGPT-ex",
"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
- Manualtrigger
- Openai
- Code
- Httprequest
- Stickynote
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
- 在n8n中导入工作流JSON文件
- 配置Gmail节点的认证信息
- 设置AI分类器的API密钥
- 自定义分类规则和标签映射
- 测试工作流执行
- 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