
Two Tools Help AI Find and Read Web Pages
Vercel's two tools let AI find web pages and read their contents.
Vercel AI Gateway now offers DeepSeek Search and DeepSeek Fetch. The two tools do different jobs when AI needs web information. Search looks for pages and returns their titles, links, and short descriptions. Fetch then gets the text from a page for the AI. Finding a page and reading its contents are separate steps. Teams can use both tools together to support the same task.
AI systems that can use tools can work with both. Teams do not need to open a separate DeepSeek account. They use one AI Gateway API key to access the tools. An API key is a code that gives software access to a service. They can keep the tools when they change AI services. This separates the choice of AI from the choice of search tools.
If DeepSeek Search fails, teams can offer other ways to search. Fetch can read a page found by any of those tools. But the AI decides how to try another search option. Teams should not expect a fixed order every time it fails. Limits on how often it tries again are also not guaranteed. The article describes these choices as attempts rather than firm promises.
Each tool costs $0.75 for every 1,000 requests that are charged. Failed requests, empty results, and requests that take too long have no tool charge. Teams also pay separately for the text the AI system processes. The search tools work with supported versions of Vercel's AI SDK. This software helps developers connect AI features inside their applications. The announcement brings web search and page reading into that existing setup.
핵심 표현
look for
~을 찾다
The tool can look for useful web pages.
try again
다시 시도하다
The system can try again after a failed search.
pay separately
별도로 비용을 내다
We pay separately for the extra service.
AI를 사용해 생성되었으며, AI는 실수를 할 수 있습니다.