Rerank

Use rerank when you already have candidate documents and want the model to score them for a query.

Examples/Sources/Rerank/main.swift

import SwiftAISDK

@main
struct RerankExample {
    static func main() async throws {
        let provider = try AIProviders.cohere()
        let model = try provider.rerankingModel("rerank-v3.5")

        let result = try await model.rerank(
            query: "Which document explains provider options?",
            documents: [
                "Provider options carry provider-specific settings.",
                "Streaming text emits lifecycle parts.",
                "Embeddings convert strings into vectors.",
            ],
            topK: 2
        )

        for item in result.results {
            print("index: \(item.index), score: \(item.score)")
        }
    }
}

Use the request initializer when your provider supports structured document objects.

let request = RerankingRequest(
    query: "Find Swift examples",
    documents: [
        ["title": "Quickstart", "body": "Generate text with a provider."],
        ["title": "Tools", "body": "Execute typed Swift tools."],
    ],
    topK: 1
)

let result = try await AI.rerank(model: model, request: request)

Options

• query is the search or ranking intent.
• documents are candidate strings or structured document objects.
• topK limits the returned matches.
• providerOptions carries provider-specific scoring settings.
• extraBody, headers, abortSignal, retryPolicy, and telemetry follow the shared facade behavior.

Results

RerankingResult returns scored document references:

• results
• warnings
• providerMetadata
• requestMetadata
• responseMetadata