Search the web and return titles, URLs, and text snippets for matching pages.

Uses DuckDuckGo as the search backend — no API key required. Results are suitable for driving follow-up fetches with WebFetch.

Args: params (WebSearchInput): Validated search parameters containing: - query (str): Search query, 1–500 characters, must not be whitespace-only. - max_results (int): Maximum results to return, 1–20 (default: 5). - region (str | None): Locale code such as "us-en" or "br-pt". When omitted the backend's worldwide default ("wt-wt") is used.

Returns: str: JSON-formatted string on success:

    {
        "results": [
            {
                "title": str,    # Page title
                "url": str,      # Full URL of the result
                "snippet": str   # Short excerpt from the page
            },
            ...
        ]
    }

    Or an error string on failure:

    "Error: <human-readable reason>"

Examples: - "Python asyncio tutorial" -> returns up to 5 results about asyncio - query="news", region="br-pt" -> returns Brazilian Portuguese news results - query=" " -> rejected by input validation before any search is performed

Error cases: - Whitespace-only or empty query: rejected by Pydantic validation - max_results outside 1–20: rejected by Pydantic validation - Backend failure (network error, rate limit, DuckDuckGo unavailable): returns "Error: Search failed: "