web-search

To add real-time web search to a backend application, instantiate the ZAI backend and invoke the web_search function with a query string to retrieve current results. This returns structured data containing URLs, snippets, and metadata for immediate processing.

Structured web search results return an array of items containing url, name, snippet, host_name, rank, date, and favicon fields. This metadata allows applications to display comprehensive search outcomes with source attribution and relevance ranking.

You can limit web search results by passing optional parameters like num for the number of items and recency_days for time filtering. These parameters allow precise control over the volume and freshness of the retrieved data.

The web search SDK supports both CLI and SDK usage for quick experiments and production integration. You can use the command line interface to rapidly test search queries before implementing them in your backend application code.

The best way to retrieve up-to-date information is using a dedicated web search SDK that returns structured results with URLs and snippets. This approach avoids building custom scrapers and provides reliable, formatted data for content discovery.

You need the z-ai-web-dev-sdk dependency installed in your backend code to execute web searches. This SDK provides the built-in search function that queries the web and returns the structured SearchFunctionResultItem array.