Class JinaEmbeddingFilter

All Implemented Interfaces:
Closeable, Serializable, AutoCloseable

public class JinaEmbeddingFilter extends OpenAIEmbeddingFilter
Extends OpenAIEmbeddingFilter for Jina AI v5 text embeddings.

The only difference from the standard OpenAI format is an optional "task" field in the request body that instructs the Jina model how to optimise the embedding. Supported values include retrieval.passage (default, for indexing documents), retrieval.query (for query-time embeddings), text-matching, classification, and separation.

Configuration key: "jina-embedding-filter"

Since:
Apache Tika 4.0
See Also:
  • Constructor Details

    • JinaEmbeddingFilter

      public JinaEmbeddingFilter()
    • JinaEmbeddingFilter

      public JinaEmbeddingFilter(InferenceConfig config)
  • Method Details

    • getTask

      public String getTask()
    • setTask

      public void setTask(String task)
      Set the Jina task type. Default is retrieval.passage. Use retrieval.query when embedding search queries.