The 10,000-document search window

Lucene search is backed by Elasticsearch, which caps its result window at 10,000 documents (offset + limit ≤ 10000). This is a hard server-side limit, not an SDK choice, and it shapes how pagination works.

What the SDK does

search.query returns a SearchPage. The page's total may be far larger than 10,000 — it is the true match count — but you can only page through the first 10,000:

with Vulners() as v:
    page = v.search.query("type:cve", limit=100)
    print(page.total)          # e.g. 250000 — the full match count
    for bulletin in page:      # iterates only up to the 10,000th document
        ...

Iterating a page (or using iter_query) transparently fetches subsequent pages and stops cleanly at the window. You never get a partial/garbage page at the boundary.

When you cross the line

Asking for data past the window fails loudly rather than silently truncating:

• search.query(query, offset=10000) raises SearchWindowExceeded before sending a request.
• Calling next_page() on a page whose next page would cross the window raises SearchWindowExceeded.

SearchWindowExceeded also subclasses ValueError, so existing except ValueError handlers around pagination keep working.

from vulners import Vulners, SearchWindowExceeded

with Vulners() as v:
    try:
        v.search.query("type:cve", offset=10001)
    except SearchWindowExceeded as err:
        print(err)   # clear message pointing you at the archive API

Retrieving more than 10,000 documents

To process an entire collection, don't paginate search — use the archive API. archive.aiter_collection("cve") streams every document in a collection with no window limit, which is the right tool for mirrors and bulk pipelines.