pocketflow-node-search

PocketFlow-Node Search Agent

A web search agent that can search the internet and provide answers based on current information using PocketFlow-Node.

What it does

This search agent can:

• Search the web for current information
• Extract relevant content from search results
• Generate comprehensive answers based on search results
• Handle multiple search queries in a single session
• Provide source citations for information

Architecture

The search agent uses a three-node flow:

1. SearchNode: Performs web searches using a search API
2. ExtractNode: Extracts and summarizes relevant content
3. AnswerNode: Generates comprehensive answers with citations

flowchart LR
    A[SearchNode] --> B[ExtractNode]
    B --> C[AnswerNode]

Loading

Features

• Web Search: Uses DuckDuckGo API for web searches
• Content Extraction: Extracts relevant information from search results
• Answer Generation: Creates comprehensive answers with citations
• Error Handling: Graceful fallbacks for search failures
• Rate Limiting: Built-in rate limiting to respect API limits

Quick Start

1. Install dependencies:

   npm install

2. Set up environment:

   cp env.example .env
   # Edit .env and add your OpenAI API key

3. Run the search agent:

   npm start

4. Start searching:

   Search: What is the latest news about AI?
   Agent: Based on recent search results, here are the latest developments in AI...
   
   Search: How does quantum computing work?
   Agent: Quantum computing is a revolutionary technology that...
   

Usage Examples

Current Events

Search: What happened in the stock market today?
Agent: Based on today's market data, the major indices showed mixed results...

Technical Information

Search: What is the difference between React and Vue?
Agent: React and Vue are both popular JavaScript frameworks, but they differ in...

Fact Checking

Search: Is it true that coffee helps with weight loss?
Agent: According to recent research, coffee may have some effects on metabolism...

Configuration

Environment Variables

• OPENAI_API_KEY: Your OpenAI API key (required)
• OPENAI_MODEL: Model to use (default: "gpt-3.5-turbo")
• SEARCH_RESULTS_LIMIT: Number of search results to process (default: 5)

Customization

You can modify the search behavior by editing:

• src/nodes.ts: Change how searches are performed and answers generated
• src/flow.ts: Modify the search flow
• src/utils/search.ts: Switch to a different search provider

Project Structure

pocketflow-node-search/
├── README.md              # This file
├── package.json           # Dependencies and scripts
├── env.example           # Environment variables template
├── src/
│   ├── main.ts           # Entry point
│   ├── nodes.ts          # Node definitions
│   ├── flow.ts           # Flow orchestration
│   └── utils/
│       ├── llm.ts        # LLM utility functions
│       └── search.ts     # Search API utilities
└── data/
    └── sample_queries.json # Sample search queries

API Reference

SearchNode

• Purpose: Performs web searches
• Input: Search query from user
• Output: Raw search results

ExtractNode

• Purpose: Extracts relevant content from search results
• Input: Raw search results
• Output: Summarized relevant content

AnswerNode

• Purpose: Generates comprehensive answers
• Input: Extracted content and original query
• Output: Final answer with citations

Troubleshooting

Common Issues

1. "Search API rate limit exceeded"

   • The agent includes rate limiting, but you may need to wait between searches

2. "No search results found"

   • Try rephrasing your query or using different keywords

3. "API key not found"

   • Make sure you've set OPENAI_API_KEY in your .env file

Debug Mode

Run with debug logging:

DEBUG=true npm start

Extending the Search Agent

Adding New Features

1. Multiple search engines: Add support for Google, Bing, or other search APIs
2. Image search: Extend to search for and analyze images
3. News filtering: Add filters for recent news vs. general information
4. Language support: Add support for searching in different languages

Example: Adding News Search

// In src/nodes.ts
class NewsSearchNode extends Node {
  async exec(query: string): Promise<any> {
    // Use a news-specific API
    const newsResults = await searchNews(query);
    return newsResults;
  }
}

License

This example is provided under the MIT license.