Run up to 10 queries across up to 5 engines in one call, merged with reciprocal-rank fusion.
Batch search fans a set of queries out to multiple engines in parallel, merges the per-engine result lists with reciprocal-rank fusion (RRF), and returns one deduplicated result set.
Each result carries the list of engines that surfaced it and a fusion score, so items found by several engines rank higher. Per-engine stats (ok flag, result count, latency) are returned alongside the merged list.
Batch responses are cached — an identical batch within the cache window returns instantly at cached-credit pricing.
All fields except queries are optional. When engines is omitted, a default set of three engines is used.
| Name | Type | Description |
|---|---|---|
queriesrequired | string[] | 1–10 search queries (max 500 chars each). |
engines | string[] | 2–5 engine ids. Defaults to `bing`, `duckduckgo`, `brave`. |
pages | number | Pages per engine, 1–5. Default: 1. |
cURL example:
A live batch costs the BATCH_SEARCH rate; a cached batch costs 1 credit. Failed engines do not block the batch — the response reports them with ok: false while successful engines still return results.