Struct PendingRerankRequest 

pub struct PendingRerankRequest { /* private fields */ }

The fluent, chainable way to build and run a rerank request.

Get one of these from Registry::rerank, add one or more with_document calls, then generate.

Example

use llmprism::Registry;

let registry = Registry::from_env();
let response = registry
    .rerank("voyageai", "rerank-2.5", "What's the capital of France?")?
    .with_document("Paris is the capital of France.")
    .with_document("Berlin is the capital of Germany.")
    .generate()
    .await?;

// Sorted most-relevant first -- `results[0]` is the best match.
println!("best match was document {}", response.results[0].index);

Implementations

impl PendingRerankRequest

pub fn new( provider: Arc<dyn Provider>, model: impl Into<String>, query: impl Into<String>, ) -> Self

Starts a new builder for provider, targeting model with the given query. You’ll normally get one of these from Registry::rerank rather than calling this directly.

pub fn with_document(self, document: impl Into<String>) -> Self

Adds one document to score. Call this more than once to rerank several documents in a single request.

pub fn with_documents( self, documents: impl IntoIterator<Item = impl Into<String>>, ) -> Self

Adds several documents to score at once.

pub fn with_top_k(self, top_k: u32) -> Self

Keeps only the top_k most relevant documents in the response, instead of every document just sorted.

pub fn with_return_documents(self, return_documents: bool) -> Self

Asks the provider to echo each document’s own text back in the response. See RerankRequest::return_documents for why you’d bother, given you already have the text locally.

pub fn to_request(&self) -> RerankRequest

Freezes the builder’s current state into a RerankRequest without sending it.

pub async fn generate(self) -> Result<RerankResponse, Error>

Sends the request and returns every document, scored and sorted by relevance (most relevant first).