React Agent for Web Search and Page Scraping

How it Works / How to Build It

• AgentNode: This is the main node that processes user input and generates responses. It uses a prompt to instruct the AI to act as a helpful support bot.
• HasFunctionCalls: This node checks if the agent's response includes any function calls (like web searches or page scrapes). It routes the workflow based on whether function calls are present.
• InvokeFunctions: If function calls are detected, this node executes them. It can handle different types of function calls, such as google_search or page_scrape.
• AccumulateChatHistory: This node collects the chat history, including the assistant's messages and results from any invoked functions, to maintain context for the conversation.
• FinalOutput: This node formats the final response from the agent, including any information gathered from the web.

What You Can Use This For

• Customer support teams can use this to provide instant answers to user queries by searching the web.
• Research teams can gather information from various sources quickly.
• Content creators can find relevant data and citations for their work.

Prerequisites

• Vellum account
• Access to web scraping and search APIs (e.g., SERP API)
• Basic understanding of how to set up workflows in Vellum

How to Set It Up

• Clone the workflow template in your Vellum account.
• Configure the AgentNode with your desired prompt settings.
• Set up the InvokeFunctions node with the necessary API keys for web scraping and search.
• Connect the AccumulateChatHistory node to ensure chat context is maintained.
• Test the workflow by inputting sample queries and verifying the responses.