Package com.redis.vl.utils.rerank

Class CohereReranker

java.lang.Object
com.redis.vl.utils.rerank.BaseReranker
com.redis.vl.utils.rerank.CohereReranker
public class CohereReranker extends BaseReranker
Reranker that uses Cohere's Rerank API to rerank documents based on query relevance.

This reranker interacts with Cohere's /rerank API, requiring an API key for authentication. The API key can be provided directly in the apiConfig Map or through the COHERE_API_KEY environment variable.

Users must obtain an API key from https://dashboard.cohere.com/. Additionally, the com.cohere:cohere-java library must be available on the classpath.

Example usage: 


 // Initialize with API key
 Map<String, String> apiConfig = Map.of("api_key", "your-api-key");
 CohereReranker reranker = CohereReranker.builder()
     .apiConfig(apiConfig)
     .limit(3)
     .build();

 // Rerank string documents
 List<String> docs = Arrays.asList("doc1", "doc2", "doc3");
 RerankResult result = reranker.rank("query", docs);

 // Rerank dict documents with rankBy fields
 List<Map<String, Object>> dictDocs = Arrays.asList(
     Map.of("content", "doc1", "source", "wiki"),
     Map.of("content", "doc2", "source", "textbook")
 );
 CohereReranker reranker2 = CohereReranker.builder()
     .apiConfig(apiConfig)
     .rankBy(Arrays.asList("content", "source"))
     .build();
 RerankResult result2 = reranker2.rank("query", dictDocs);
See Also:

  • Method Details

    • rank

      public RerankResult rank(String query, List<?> docs)
      Rerank documents based on query relevance using Cohere's Rerank API.
      Specified by:
      rank in class BaseReranker
      Parameters:
      query - The search query
      docs - List of documents (either List<String> or List<Map<String, Object>>)
      Returns:
      RerankResult with reranked documents and relevance scores
      Throws:
      IllegalArgumentException - if query or docs are invalid

    • rank

      public RerankResult rank(String query, List<?> docs, Map<String,Object> kwargs)
      Rerank documents based on query relevance using Cohere's Rerank API with runtime parameter overrides.
      Parameters:
      query - The search query
      docs - List of documents (either List<String> or List<Map<String, Object>>)
      kwargs - Optional parameters to override defaults (limit, return_score, rank_by, max_chunks_per_doc)
      Returns:
      RerankResult with reranked documents and relevance scores
      Throws:
      IllegalArgumentException - if query or docs are invalid