web-search

Import ZAI from z-ai-web-dev-sdk, create an instance with ZAI.create(), then call zai.functions.invoke with the web_search function name, passing a query and optional result count. Results return as an array of items with URL, title, snippet, and date.

Use the z-ai CLI with the function command: z-ai function -n web_search -a '{"query": "your topic", "num": 5}'. Add the -o flag with a file path to save results as JSON, and use recency_days to filter by date.

No, the z-ai-web-dev-sdk must only be used in backend code. Importing it in client-side code will fail and could expose credentials. Call it from server-side routes or API endpoints instead.

Each result is a SearchFunctionResultItem containing url, name (page title), snippet (preview text), host_name (domain), rank, date, and favicon. You can map these fields into custom formats for display or further processing.

Empty results usually come from overly vague query terms, connectivity issues, or API problems. Refine the query with specific keywords, verify the response is an array, and implement retry logic with exponential backoff for transient failures.

Implement a rate limiter that tracks request timestamps and waits when the per-minute threshold is reached. Combine this with result caching using a Map keyed by query, so repeated searches within the cache window skip the API call entirely.