Hello everyone!
How can I implement a similar functionality as in chat-gpt web-search button? I will collect some input data that I need to search on the internet and generate a detailed description based on the information I find.
I’ve seen options involving SERP, but I’m not sure if it’s the most best-practice or new way to go.
To build a ChatGPT-like web search feature, you’ll need two main components: (1) a search API to fetch real-time data, and (2) an LLM agent that decides when and how to search, then summarizes the results. A modern, LLM-friendly choice for search is Tavily because it returns clean, concise JSON optimized for AI processing. You could also use Google Custom Search or Bing Search for broader coverage, while SERP APIs (e.g., SerpAPI) remain viable but can be pricier.
Using an agent framework like LangGraph or LangChain, you can set up a create_react_agent that invokes the search tool automatically based on user intent, retrieves key points from the results, and crafts a detailed response. This mirrors how ChatGPT’s web search button works: user input → detect search need → query API → extract facts → synthesize final answer.