Package com.tnsai.intelligence.rag

Class HybridRAGStrategy

java.lang.Object

com.tnsai.intelligence.rag.HybridRAGStrategy

All Implemented Interfaces:

RAGStrategy

public final class HybridRAGStrategy
extends Object
implements RAGStrategy

RAG strategy that fuses BM25, vector similarity, and — when a GraphRAGStrategy is supplied — knowledge-graph hits via Reciprocal Rank Fusion (RRF).

The strategy name stays hybrid. A missing graph capability does not degrade to VECTOR-only: BM25 and vector still fuse, and graphStream() reports HybridGraphStream.SKIPPED_NO_GRAPH_STORE. Unweighted RRF uses ReciprocalRankFusion.DEFAULT_K (60).

Two-stream construction still delegates to HybridMemoryRetriever so existing lexical/semantic scores stay unchanged. The three-stream constructor ranks each source independently, then fuses the lists. Scores are normalized to 0.0..1.0 against the number of non-empty streams before RAGContext.minScore() applies.

Since:

0.2.4

Field Summary

Fields

Modifier and Type	Field	Description
static final String	METADATA_GRAPH_STREAM	Metadata key for HybridGraphStream.metadataValue().
static final String	METADATA_STREAMS	Metadata key listing the streams that contributed to a hit.
static final String	STREAM_BM25	Provenance token for the BM25 stream.
static final String	STREAM_GRAPH	Provenance token for the graph stream.
static final String	STREAM_VECTOR	Provenance token for the vector stream.

Constructor Summary

Constructors

Constructor	Description
HybridRAGStrategy(com.tnsai.memory.advanced.BM25Index bm25Index, com.tnsai.memory.advanced.VectorMemoryStore vectorStore)	Creates a two-stream hybrid strategy from BM25 and vector stores.
HybridRAGStrategy(com.tnsai.memory.advanced.BM25Index bm25Index, com.tnsai.memory.advanced.VectorMemoryStore vectorStore, GraphRAGStrategy graphStrategy)	Creates a three-stream hybrid strategy that RRF-fuses BM25, vector, and graph rankings.
HybridRAGStrategy(com.tnsai.memory.advanced.HybridMemoryRetriever hybridRetriever)	Creates a two-stream hybrid strategy with a pre-configured retriever.

Method Summary

Modifier and Type	Method	Description
HybridGraphStream	graphStream()	How the optional graph stream was handled for this instance.
String	name()	Returns the unique name of this strategy (e.g., "vector", "keyword", "hybrid").
List<RetrievedDocument>	retrieve(String query, RAGContext context)	Retrieves documents relevant to the given query.

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

METADATA_STREAMS

public static final String METADATA_STREAMS

Metadata key listing the streams that contributed to a hit.

See Also:

• Constant Field Values

METADATA_GRAPH_STREAM

public static final String METADATA_GRAPH_STREAM

Metadata key for HybridGraphStream.metadataValue().

See Also:

• Constant Field Values

STREAM_BM25

public static final String STREAM_BM25

Provenance token for the BM25 stream.

See Also:

• Constant Field Values

STREAM_VECTOR

public static final String STREAM_VECTOR

Provenance token for the vector stream.

See Also:

• Constant Field Values

STREAM_GRAPH

public static final String STREAM_GRAPH

Provenance token for the graph stream.

See Also:

• Constant Field Values

Constructor Details

HybridRAGStrategy

public HybridRAGStrategy(com.tnsai.memory.advanced.BM25Index bm25Index,
 com.tnsai.memory.advanced.VectorMemoryStore vectorStore)

Creates a two-stream hybrid strategy from BM25 and vector stores.

Uses the default RRF constant (k=60). The graph stream is skipped with HybridGraphStream.SKIPPED_NO_GRAPH_STORE.

Parameters:

bm25Index - the BM25 keyword index

vectorStore - the vector similarity store

HybridRAGStrategy

public HybridRAGStrategy(com.tnsai.memory.advanced.HybridMemoryRetriever hybridRetriever)

Creates a two-stream hybrid strategy with a pre-configured retriever.

Parameters:

hybridRetriever - the hybrid retriever to use

HybridRAGStrategy

public HybridRAGStrategy(com.tnsai.memory.advanced.BM25Index bm25Index,
 com.tnsai.memory.advanced.VectorMemoryStore vectorStore,
 GraphRAGStrategy graphStrategy)

Creates a three-stream hybrid strategy that RRF-fuses BM25, vector, and graph rankings.

Parameters:

bm25Index - the BM25 keyword index

vectorStore - the vector similarity store

graphStrategy - source-fenced graph traversal used as the third list

Since:

0.14.0

Method Details

graphStream

public HybridGraphStream graphStream()

How the optional graph stream was handled for this instance.

Returns:

fused or skipped-without-store

Since:

0.14.0

name

public String name()

Description copied from interface: RAGStrategy

Returns the unique name of this strategy (e.g., "vector", "keyword", "hybrid").

Names are compared case-insensitively using Locale.ROOT.

Specified by:

name in interface RAGStrategy

Returns:

the strategy name, never null

retrieve

public List<RetrievedDocument> retrieve(String query,
 RAGContext context)

Description copied from interface: RAGStrategy

Retrieves documents relevant to the given query.

Every returned document must exactly match all metadata filters in the context. Implementations apply that constraint before limiting the result to RAGContext.maxResults().

Specified by:

retrieve in interface RAGStrategy

Parameters:

query - the search query (never null)

context - retrieval parameters (max results, min score, exact metadata filters)

Returns:

an unmodifiable list of retrieved documents, sorted by descending relevance