SearchStrategy

Search strategy for creating search requests during prompt augmentation.

This is a functional interface (SAM) that defines how a user query string should be transformed into a SearchRequest for retrieval.

SimilaritySearchStrategy is the default implementation. It uses vector embeddings for semantic search and works with all supported vector backends.

Pre-built implementations are available for common search types:

• SimilaritySearchStrategy - Vector similarity search (semantic search)

Usage Examples

Using pre-built strategies (Kotlin):

retrieval {
    searchStrategy = SimilaritySearchStrategy(topK = 5, similarityThreshold = 0.7)
}

Custom implementation as lambda (Java):

SearchStrategy customStrategy = (query) ->
    new SimilaritySearchRequest(query, 5, 0, 0.8, null);

Inheritors