Skip to main content
Version: Winter 2024 SheCodes Program

Coze Workflow Development

🚀 Platform Introduction​

Coze is a powerful AI workflow development platform. Through this tutorial, you will learn how to build a complete AI-driven workflow.

Tutorial Videos​

🎥 Environment Setup Tutorial 1

🎥 Environment Setup Tutorial 2

🎥 Environment Setup Tutorial 3

🎯 Learning Objectives​

Through this tutorial, you will master:

  • ✅ Creating complete Coze workflows
  • ✅ Implementing automatic webpage content retrieval
  • ✅ Using AI to generate high-quality summaries
  • ✅ Saving data to Notion

🛠️ Environment Preparation​

Notion Configuration​

1. Create Integration​

  1. Visit Notion Developers
  2. Click "Create new integration"
  3. Fill in integration information:
    • Name: Article Summary
    • Associated workspace: Select your workspace
  4. Save the generated Integration Token

2. Get Page ID​

// Notion page URL format
https://notion.so/workspace/page-title-32-character-string

// Extract 32-character string as page_id
const pageId = url.match(/[a-f\d]{32}/)[0];

Coze Configuration​

1. Create Project​

  • Visit Coze website and login
  • Create new workflow project
  • Configure basic project information

2. Required Nodes​

  • 🌐 Browse Node - Web scraping
  • 🤖 LLM Node - Generate summaries
  • 📝 Text Processing Node - Text processing
  • 💾 Write Page Node - Notion integration

🔧 Workflow Construction​

1. Browse Node Configuration​

Node Function​

  • Visit specified URL
  • Extract webpage content
  • Handle different types of webpages

Parameter Settings​

// url parameter configuration
{
  "url": "{{input}}",
  "enforce_crawl": true
}

Output Structure​

{
  "code": 200,
  "message": "success",
  "plugin_observation": {
    "observation": "Actual webpage content",
    "natural_language_desc": "Content description information"
  }
}

2. LLM Node Configuration​

System Prompt​

You are a professional article summarizer. Your task is to read the provided article and create a CONCISE summary that:
- Captures only the most essential points (max 2-3 key points)
- Uses very concise language
- Keeps the total output under 1500 characters
- Maintains clarity while being brief
- Includes a short title

Format your response as:
Title: [Short Article Title]
Summary: [Your concise summary in 1-2 paragraphs]
Key Points:
- [Point 1]
- [Point 2]
- [Point 3]

User Prompt​

Please provide a summary of the following article:
{{browse.plugin_observation.observation}}

3. Text Processing Node Configuration​

Input Parameters​

{
  "String1": "{{llm.output.output}}"
}

Text Template​

=========================================
                Article Summary
=========================================

{{String1}}
=========================================

4. Write Page Node Configuration​

Notion Parameters​

{
  "page_id": "Your Notion page ID",
  "text": "{{Text Processing.output}}"
}

🎨 Optimization Strategies​

Prompt Optimization​

  • Use clear role definitions
  • Set specific output requirements
  • Maintain prompt consistency
  • Regularly update and optimize prompts

Performance Monitoring​

Response Time Monitoring​

const metrics = {
  browse_time: 'Scraping time',
  llm_time: 'AI processing time',
  storage_time: 'Storage time',
  total_time: 'Total time'
};

Optimization Goals​

  • Web scraping < 3 seconds
  • AI processing < 5 seconds
  • Data storage < 2 seconds

🚫 Common Issues​

Network Issues​

  • URL Inaccessible: Check URL format and network connection
  • Incomplete Content Retrieval: Adjust enforce_crawl settings
  • Authorization Verification Failed: Check Token validity

Processing Issues​

  • Format Parsing Error: Optimize prompt settings
  • Character Encoding Issue: Adjust text processing parameters
  • Content Truncation Error: Increase text length limit

🎯 Testing and Optimization​

Functional Testing​

Input Validation​

  • Test different URL formats
  • Verify special character handling
  • Check error handling mechanisms
  • Test boundary conditions

Process Validation​

  • Node connection correctness
  • Data transfer completeness
  • Exception handling effectiveness
  • Output format standardization

Performance Testing​

Concurrency Testing​

  • Multiple URL simultaneous processing
  • Batch task execution
  • High-frequency request handling

Resource Monitoring​

  • Memory usage
  • CPU utilization
  • API call frequency

🌟 Best Practices​

Daily Maintenance​

  1. Regularly check functional completeness
  2. Monitor error logs
  3. Update and optimize prompts
  4. Adjust processing strategies

Continuous Optimization​

  1. Collect user feedback
  2. Analyze usage data
  3. Optimize processing flow
  4. Expand feature capabilities

🔄 Development Process​

Workflow Features
  • Automated article summary generation
  • Structured data storage solution
  • Scalable system architecture
  • Efficient processing flow

Through this Coze workflow project, you will master how to build a complete AI-driven workflow, from web scraping to intelligent summary generation, to data storage - the full development process.