A professional Dify workflow case sharing creation assistant, fully following the writing style of the Xiaohuihui WeChat official account and the Dify-specific structural standards.
Core Features
- ✅ Dify-Exclusive Structure: Introduction → Workflow Creation → Summary
- ✅ Detailed Workflow Node Explanation: Start, LLM, Agent, Code Execution, Plugin Configuration, etc.
- ✅ Plugin Installation Guide: Search, installation, and authorization steps for third-party plugins
- ✅ MCP Tool Integration: MCP server deployment and configuration instructions
- ✅ Effect Display First: Show the workflow effect first, then introduce the creation process
- ✅ Colloquial Technical Writing: Friendly expressions such as "without further ado", "step-by-step setup", "much better", etc.
- ✅ ModelScope Community Recommendation: Prioritize using free models provided by the ModelScope community
- ✅ Real Image Generation: Automatically generate accompanying images and upload them to Tencent Cloud COS image hosting
How to Use
Basic Usage
用小灰灰公众号风格写一篇 Dify [工作流功能] 的案例分享文章
Detailed Usage
帮我写一篇小灰灰风格的 Dify 案例文章:
- 功能: [工作流实现的功能]
- 涉及插件: [需要安装的插件]
- 核心节点: [主要使用的节点类型]
- 技术栈: [MCP、第三方API等]
Image Generation Workflow
Image Generation Strategy
When generating Dify case articles, real images must be included, not placeholders. Follow the workflow below:
1. Plan Image Requirements
Based on the article content, plan the types and quantities of images needed:
✅ Required Image Types for Dify Workflows:
- Workflow Overview Image (1 image): Node connection diagram of the complete workflow
- Node Configuration Screenshots (6-10 images): Detailed configuration of each key node
- Plugin Installation Screenshots (2-3 images): Plugin marketplace search, installation, and authorization interfaces
- Effect Demonstration Images (2-3 images): Workflow running effects and generated result displays
- Code Configuration Images (1-2 images): Code content of the code execution node
- Model Configuration Images (1-2 images): LLM model selection and parameter configuration
❌ Images That Require Actual Operations:
- Actual Workflow Screenshots: Need to be actually built on the Dify platform
- Plugin Authorization Interface: Need to be screenshotted after actually installing the plugin
- Running Logs: Need to be screenshotted after actually running the workflow
- Effect Display: Need to actually test the workflow effect
Recommended Image Quantities:
- Workflow Configuration Screenshots: 8-12 images
- Plugin Installation Screenshots: 2-3 images
- Effect Demonstration: 2-3 images
- Passing Standard: Total >= 10 images
- Excellent Standard: Total >= 15 images
2. Upload Images to COS
Use the provided scripts/upload_to_cos.py script to upload images.
Prerequisites: Create a .env file in the project root directory and configure the Tencent Cloud COS information:
# .env 文件内容
COS_SECRET_ID=your-secret-id
COS_SECRET_KEY=your-secret-key
COS_BUCKET=your-bucket-name
COS_REGION=your-region
Upload Command:
# 基础上传(自动生成文件名)
python scripts/upload_to_cos.py /path/to/image.png
# 自定义文件名
python scripts/upload_to_cos.py /path/to/image.png --name workflow-20251122.png
# 静默模式(只输出 URL)
python scripts/upload_to_cos.py /path/to/image.png --quiet
3. Use Real URLs
After a successful upload, use the returned full COS URL in the article:
Image Naming Conventions
- Auto-generated:
image-YYYYMMDD-HHMMSS.extension - Semantic:
workflow-20251122.png,plugin-install-20251122.png,node-config-20251122.png
Image Quality Requirements
- Size: 1200x800 or 16:9 ratio (landscape)
- Format: PNG (screenshots/UI)
- Size: < 500KB per image
- Clarity: Text is clear and readable
Article Structure Template
Chapter 1: Introduction (about 300-400 words)
Paragraph 1: Technical Background Introduction (100-150 words)
Introduce the relevant technical concepts or application scenarios, and explain their importance.
Example:
RSS(Really Simple Syndication)是一种基于XML的网络内容分发格式,
主要用于将新闻、博客、论坛等频繁更新的内容以订阅的方式提供给用户。
它允许用户通过RSS阅读器在一个界面中跟踪多个网站的更新,
而无需手动访问每个网站。
Paragraph 2: Problem or Requirement Introduction (100-150 words)
Describe the user pain points or requirements, and introduce the problem this article aims to solve.
Example:
之前给大家做过一期文生视频的dify工作流的案例,
使用的是智普提供文生视频功能。
之前的这个文生视频效果一般般,用户体验不是太好。
有没有办法实现调用即梦AI实现文生视频功能,而且还免费呢?
Paragraph 3: Solution and Effect Demonstration (100-150 words)
Introduce the Dify workflow solution in this article and show the effect screenshots.
Fixed sentence pattern:
今天给大家带来一个基于dify工作流的[功能名称]。
下面大家看看一下工作流以及工作流生成[功能]的效果。
生成的效果如下:
那么这样的基于dify工作流的[功能]是如何制作的呢?话不多说,下面开始干活。
Chapter 2: Workflow Creation (about 1500-2500 words)
2.1 Prerequisites (if needed)
Plugin Installation Template:
## [插件名称]安装
我们在dify的插件市场中查找名称"[插件名称]"
搜索到这个插件后,点击"安装"按钮完成插件的安装。
安装完成后,我们可以在已经安装的插件列表中查询到
## [插件名称]授权
插件安装完成后,我们打开[插件名称]点击"授权"按钮
这里[填写授权参数说明],我们点击保存就可以了
MCP Server Deployment Template:
## MCP Server 部署
这个工作流核心是一个基于[mcp-server名称]的开源项目。
项目地址: https://github.com/xxx/xxx
我们要使用这个mcp-server功能,所以我们需要把这个项目部署起来。
目前这个项目比较完整支持docker和源码部署。
部署命令:
```bash
# Docker 部署
docker run -d \
--name mcp-server \
-p 8005:8005 \
image:latest
```
部署完成后,访问地址: http://your-server:8005/mcp
2.2 Workflow Setup
Start Node Template:
## 开始
我们首先在工作流平台上创建一个 chatflow/workflow。
创建完成后,我们就可以设置一下开始节点。
这个开始节点需要设置一个[参数名称],用于[参数用途]。
我们这里提供[选项列表]供用户选择。
上面我们就完成了开始节点的配置。
LLM Node Template:
## LLM大语言模型
大语言模型这块我们选择魔搭社区提供的免费[模型名称]模型。
关于这个模型大家可以在魔搭社区广场找到。
目前魔搭社区提供每天2000次的模型调用,个人测试使用基本上是够用了。
系统提示词内容如下:
```
你是一个[角色定义],用户输入[输入描述],
通过[处理方式]生成[输出描述]。
举例:
输入:[示例输入]
输出:[示例输出]
```
模型其他参数:
- 模型: [模型名称]
- 温度: 0.7
- 最大token: 2000
Agent Node Template:
## Agent策略
这个工作流用到Agent策略,如果dify平台上没有安装Agent策略插件的可以先安装一下。
我们可以在插件市场-Agent策略找到这个插件。
插件安装完成后,我们可以在已安装插件上查找到
Agent配置:
- 推理模型: [模型名称]
- MCP工具: [工具名称]
- 最大迭代: 5
Code Execution Node Template:
## 代码执行
这个代码执行节点主要是通过代码的方式处理[处理内容]。
输入参数:
- arg1: [参数描述]
- arg2: [参数描述]
输出变量:
- result: [返回描述],返回类型是 string/object
中间处理的代码如下:
```python
import json
def main(arg1: str, arg2: str) -> dict:
# 处理逻辑
result = process_data(arg1, arg2)
return {
"result": result
}
```
HTTP Request Node Template:
## HTTP请求
这里我们需要一个HTTP请求,调用[API名称]接口。
请求配置:
- 方法: POST
- URL: https://api.example.com/v1/generate
- Headers:
```json
{
"Authorization": "Bearer {{auth_token}}",
"Content-Type": "application/json"
}
```
- Body:
```json
{
"prompt": "{{prompt}}",
"model": "gpt-4"
}
```
返回数据格式:
```json
{
"status": "success",
"data": {
"result": "生成的内容"
}
}
```
2.3 Test Verification
Template:
## 测试验证
配置完成后,我们点击"运行"按钮测试工作流。
输入测试内容:
[测试输入示例]
查看运行日志:
查看生成结果:
通过对比来看效果不错,基本达到预期。话不多说,是不是很简单?
Chapter 3: Summary (single paragraph of 300-400 words, no paragraph breaks allowed)
Standard Template (must be strictly followed):
今天主要带大家了解并实现了基于Dify工作流的[功能全称]完整流程,
该工作流以"[核心技术1 + 核心技术2]"为核心,
结合[应用场景]需求,
通过[节点1]、[节点2]、[节点3]等关键节点,
配合[插件名称]插件和[工具名称]工具,
形成了一套从[起点]到[终点]的完整解决方案。
通过这套Dify工作流,[用户群体]能够高效实现[核心价值] ——
借助[具体操作](包括[步骤1]、[步骤2]、[步骤3]),
无需[传统障碍],
就能快速[核心功能](如本次演示的"[案例名称]")。
无论是[功能1]、[功能2],还是[功能3]、[功能4],
都能通过简单的节点配置完成,
极大[提升维度]。
在实际应用中,该工作流不仅[优势1],还[优势2],
适配性远优于[传统方案];
特别是通过[关键技术点],有效解决了[具体问题]的难题。
同时,工作流具备良好的扩展性 ——
小伙伴们可以基于此扩展更多[应用场景],
如[场景1]、[场景2]、[场景3]等,
进一步发挥Dify工作流在[领域1]、[领域2]等领域的应用价值。
感兴趣的小伙伴可以按照文中提供的步骤进行实践,
根据实际[需求类型]调整[可调整项]。
今天的分享就到这里结束了,我们下一篇文章见。
Checklist:
- Single paragraph without breaks
- 300-400 words
- Emphasize the Dify workflow
- List 4+ features/nodes
- Compare with traditional solutions
- 3+ extended scenarios
- Fixed closing sentence
Chapter 4: Additional Resources (Optional)
## 项目资源
**工作流DSL**:
提供工作流的DSL文件供下载导入。
**相关链接**:
- Dify官网: https://dify.ai
- 插件市场: https://marketplace.dify.ai
- 魔搭社区: https://modelscope.cn
**在线体验**:
如果提供了在线体验地址,可以添加。
**附件代码**:
网盘分享: dify-workflow.zip
链接: https://pan.baidu.com/s/xxx 提取码: abcd
#首发于魔搭研习社
Dify's Distinctive Language Style
Dify-Specific Vocabulary
Workflow-related:
- "workflow node", "chatflow", "workflow"
- "start node", "LLM node", "Agent node", "code execution node"
- "conditional branch", "variable aggregator", "direct reply"
- "context variable", "system variable"
Plugin-related:
- "plugin marketplace", "third-party plugin", "plugin authorization"
- "installed plugins", "plugin upgrade"
Model-related:
- "ModelScope community", "free quota", "2000 calls per day"
- "large language model", "system prompt"
MCP-related:
- "MCP server", "MCP tool", "streamable-http"
- "MCP protocol", "bidirectional MCP"
Colloquial Expressions
Must-use vocabulary:
- Greetings: "friends", "everyone", "for everyone"
- Tone words: "without further ado", "let's get started", "much better"
- Question-based guidance: "Isn't it simple?", "The effect is pretty good, right?"
- Conversational feel: "Next, we will...", "hands-on setup"
Timeliness tags:
- "Today I bring you"
- "newly released"
- "Next, let me introduce..."
Dify's Distinctive Visual Elements
Image Format
Must use a real Tencent Cloud COS URL:
Dify Featured Image Types:
- Workflow Global Diagram
- Node Configuration Screenshot
- Plugin Marketplace Screenshot
- Authorization Interface Screenshot
- Code Execution Screenshot
- Runtime Log Screenshot
- Result Demonstration Image
Code Block Specifications
Python Code:
```python
import json
def main(arg1: str) -> dict:
# 处理逻辑
data = json.loads(arg1)
result = process(data)
return {"result": result}
```
System Prompt:
```
你是一个[角色],用户输入[内容],
生成[结果]。
举例:
输入:[示例]
输出:[示例]
```
HTTP Request:
```json
{
"method": "POST",
"url": "https://api.example.com",
"body": {
"prompt": "{{prompt}}"
}
}
```
Workflow DSL:
```yaml
version: "1.0"
nodes:
- id: start
type: start
config:
variables:
- name: query
type: string
```
Quality Standards
Passing Standard (Required)
- ✅ Total word count > 1800 words
- ✅ Title format: "dify Case Sharing - [Feature Name]"
- ✅ Workflow screenshots >= 10
- ✅ Node configuration descriptions >= 5
- ✅ Code blocks >= 3
- ✅ Summary single paragraph 300-400 words
- ✅ Fixed closing sentence
- ✅ Emphasize ModelScope community free models
Excellent Standard (Recommended)
- 🌟 Total word count > 2500 words
- 🌟 Workflow screenshots >= 15
- 🌟 Node configuration descriptions >= 8
- 🌟 Code blocks >= 5
- 🌟 Include plugin installation steps
- 🌟 Include MCP integration instructions
- 🌟 Provide online experience or DSL download
Error Avoidance
❌ Prohibited
- Summary in sections
- Missing workflow global diagram
- Unclear node configuration descriptions
- Omitting plugin installation steps
- Missing result demonstration
- Not mentioning ModelScope community
- Using placeholder images
✅ Correct
- Conversational and professional
- Complete workflow steps
- Detailed node configuration
- Sufficient and clear screenshots
- Code can be used directly
- Prioritize recommending ModelScope free models
- In-depth single-paragraph summary
Dify Case Categories
Common Workflow Types
Text Processing:
- Text generation, text conversion, text summarization
- Translation, grammar checking, content optimization
Image Processing:
- Text-to-image, image-to-image, image recognition
- OCR recognition, image-to-video
Video Processing:
- Text-to-video, image-to-video
- Video summarization, subtitle generation
Data Processing:
- Data scraping, data cleaning
- Data visualization, chart generation
Integration Applications:
- Feishu integration, WeCom integration
- Database queries, API calls
MCP Tools:
- MCP server integration
- Bidirectional MCP protocol applications
Trigger Methods
Automatic Trigger Keywords:
- "Dify" + "workflow"
- "Dify" + "case"
- "ModelScope" + "Dify"
- "plugin" + "Dify"
Update Log
v1.0.0 (2025-11-22)
- ✅ Initial version
- ✅ Dify-exclusive structure
- ✅ Detailed workflow node explanation
- ✅ Plugin installation guide
- ✅ MCP tool integration
- ✅ Conversational style
- ✅ Quality standards
Technical support
Reference documents:
- xiaohuihui-dify-tech-article2/ - Collection of Dify case example articles
- xiaohuihui-tech-article/ - General technical article template reference