Skip to content

GETTING_STARTED.md

Latest commit

 

History

History
220 lines (143 loc) · 5.39 KB

GETTING_STARTED.md

File metadata and controls

220 lines (143 loc) · 5.39 KB
Raw

Getting Started with Agent Workflow Builder

Welcome to the Agent Workflow Builder! This guide will help you create your first AI-powered workflow in just a few minutes.

🚀 Quick Start

1. Installation

Windows:

# Run the setup script
.\scripts\setup.ps1

# Or manually install dependencies
npm install

macOS/Linux:

# Run the setup script
chmod +x scripts/setup.sh
./scripts/setup.sh

# Or manually install dependencies
npm install

2. Start the Application

npm start

Open your browser and go to http://localhost:3000

🎯 Your First Workflow

Let's create a simple workflow that takes text input and processes it with AI.

Step 1: Create a New Workflow

  1. Click "New Workflow" on the home screen
  2. Enter a name like "My First Workflow"
  3. Click "Create"

Step 2: Add Nodes

  1. Click "Show Node Palette" to reveal the node types
  2. Drag a Data Input node onto the canvas
  3. Drag an LLM Task node onto the canvas
  4. Drag a Data Output node onto the canvas

Step 3: Connect the Nodes

  1. Click and drag from the bottom of the Data Input node to the top of the LLM Task node
  2. Click and drag from the bottom of the LLM Task node to the top of the Data Output node

Step 4: Configure the Nodes

  1. Click on the Data Input node to open its configuration:

    • Set Data Type to "text"
    • Enter default value: "Hello, world!"

  2. Click on the LLM Task node to open its configuration:

    • Enter prompt: "Analyze the following text and provide insights: {{input}}"
    • Select model: "gpt-3.5-turbo"
    • Set temperature: 0.7

  3. Click on the Data Output node to open its configuration:

    • Set format to "text"
    • Enter filename: "analysis.txt"

Step 5: Test Your Workflow

  1. Go to the Testing Panel on the right
  2. Enter some test text in the input field
  3. Click "Run Test"
  4. Watch as each node processes the data
  5. View the results in the execution panel

Step 6: Save Your Workflow

  1. Click "Save" in the toolbar
  2. Your workflow is now saved and can be loaded later

🌟 Example Workflows

Web Scraping + AI Analysis

Create a workflow that:

  1. Takes a URL as input
  2. Scrapes the website content
  3. Analyzes the content with AI
  4. Outputs the analysis

Nodes needed:

  • Data Input (URL)
  • Web Scraping
  • LLM Task (Analysis)
  • Data Output

Document Processing Pipeline

Create a workflow that:

  1. Takes document text
  2. Generates embeddings
  3. Performs similarity search
  4. Returns similar documents

Nodes needed:

  • Data Input (Text)
  • Embedding Generator
  • Similarity Search
  • Data Output

🔧 Node Types Explained

Data Input/Output

  • Purpose: Entry and exit points for your workflow
  • Configuration: Data type, default values, output format

Web Scraping

  • Purpose: Extract content from websites
  • Configuration: URL, CSS selectors, content length limits

LLM Task

  • Purpose: Process text with AI language models
  • Configuration: Prompts, model selection, temperature, token limits

Embedding Generator

  • Purpose: Create vector representations of text
  • Configuration: Model selection, vector dimensions

Similarity Search

  • Purpose: Find similar content in vector databases
  • Configuration: Vector store, result count, similarity threshold

Structured Output

  • Purpose: Parse and structure data according to schemas
  • Configuration: JSON schema, model selection

💡 Tips & Best Practices

Workflow Design

  • Start Simple: Begin with basic workflows and add complexity gradually
  • Clear Naming: Use descriptive names for nodes and workflows
  • Test Frequently: Test your workflow at each step
  • Error Handling: Consider what happens when nodes fail

Node Configuration

  • Use Variables: Reference other node outputs with {{nodeId.output}}
  • Set Reasonable Limits: Configure appropriate token and length limits
  • Choose Right Models: Select models based on your specific needs

Performance

  • Parallel Processing: Some nodes can run in parallel
  • Caching: Consider caching results for expensive operations
  • Monitoring: Use the execution panel to monitor performance

🐛 Troubleshooting

Common Issues

"Nodes not connecting"

  • Make sure you're dragging from output handles (bottom) to input handles (top)
  • Check that nodes are properly positioned

"Workflow not executing"

  • Verify all required parameters are configured
  • Check the execution panel for error messages
  • Ensure all nodes are properly connected

"Import/Export not working"

  • Make sure you're using valid JSON format
  • Check browser console for error messages

Getting Help

  • Check the Execution Panel for detailed error messages
  • Use the Testing Panel to debug with sample data
  • Review the README.md for detailed documentation

🎉 Next Steps

Now that you've created your first workflow:

  1. Explore More Node Types: Try different AI operations
  2. Create Complex Workflows: Combine multiple node types
  3. Share Your Workflows: Export and share with others
  4. Customize the Interface: Modify node configurations
  5. Integrate with APIs: Connect to real AI services

📚 Additional Resources

Happy building! 🚀