SearchIterator()

Creates an iterator that retrieves search results in batches while preserving the server search token and one MVCC session timestamp across pages.

pub async fn search_iterator(&self, request: SearchIteratorRequest) -> Result<SearchIterator>

Request Syntax

let request = SearchIteratorRequest::builder()
    .search(
        SearchRequest::builder()
            .collection_name("books")
            .vector_field("embedding")
            .vectors(SearchVectors::Float(vec![vec![0.1, 0.2, 0.3, 0.4]]))
            .limit(1000)
            .build()?,
    )
    .batch_size(100)
    .build()?;
let mut iterator = client.search_iterator(request).await?;

REQUEST FIELDS:

• search: SearchRequest

  The SearchRequest that defines the collection, vector field, and query vectors. Search iterators do not support primary-key IDs or order_by_fields. Required.

• batch_size: usize

  Number of hits returned per next() call. Defaults to 1000.

• limit: Option<usize>

  Optional total number of hits to iterate before stopping.

RETURNS:

Result

Returns a SearchIterator. Call next() to receive the next SearchResponse batch (None when exhausted) and close() when done. Returns Error on failure.

Example

let request = SearchIteratorRequest::builder()
    .search(
        SearchRequest::builder()
            .collection_name("books")
            .vector_field("embedding")
            .vectors(SearchVectors::Float(vec![vec![0.1, 0.2, 0.3, 0.4]]))
            .limit(1000)
            .build()?,
    )
    .batch_size(100)
    .build()?;
let mut iterator = client.search_iterator(request).await?;
while let Some(batch) = iterator.next().await? {
    println!("got {} hits", batch.results().len());
}