Interface IRetrievalStrategy
Defines a strategy for retrieving matching partitions from indexed data sources.
public interface IRetrievalStrategyExamples
// Assign a retrieval strategy to a RagEngine instance.
LM embeddingModel = LM.LoadFromModelID("embeddinggemma-300m");
RagEngine ragEngine = new RagEngine(embeddingModel);
// Use the default vector (semantic) strategy.
ragEngine.RetrievalStrategy = new VectorRetrievalStrategy();
// Switch to hybrid (vector + keyword) retrieval.
ragEngine.RetrievalStrategy = new HybridRetrievalStrategy();Remarks
Implementations control how candidate partitions are scored and selected during the initial retrieval phase. Post-retrieval processing such as reranking and MMR diversity filtering is applied by RagEngine regardless of which strategy is used.
LM-Kit provides three built-in strategies: VectorRetrievalStrategy (semantic similarity), Bm25RetrievalStrategy (keyword matching), and HybridRetrievalStrategy (combined with Reciprocal Rank Fusion).
Properties
- RequiresQueryVector
Gets a value indicating whether the strategy requires a query embedding vector.
Methods
- RetrieveAsync(IReadOnlyList<DataSource>, string, float[], int, float, bool, bool, DataFilter, CancellationToken)
Retrieves matching partitions from the given data sources.