Query-centric grounding: what does the model retrieve for a query, what does it keep, and what does each snippet look like.
Runs Google's native grounded search: the model answers your query, and we capture the sources it cited (with their real URLs, resolved from Google's redirect links), which answer passages cite each source, and the retrieved content per source. A run is a single grounded call (~1 min) — it runs in the background and the results appear here when ready.