Class HybridGraphRetriever<T>

Namespace

AiDotNet.RetrievalAugmentedGeneration.Graph

Assembly

AiDotNet.dll

Hybrid retriever that combines vector similarity search with graph traversal for enhanced RAG.

public class HybridGraphRetriever<T>

Type Parameters

T

The numeric type used for vector operations.

Inheritance

object

HybridGraphRetriever<T>

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

This retriever uses a two-stage approach: 1. Vector similarity search to find initial candidate nodes 2. Graph traversal to expand context with related nodes

For Beginners: Traditional RAG uses only vector similarity:

Query: "What is photosynthesis?" Traditional RAG:

• Find documents similar to the query
• Return top-k matches
• Misses related context!

Hybrid Graph RAG:

• Find initial matches via vector similarity
• Walk the graph to find related concepts
• Example: photosynthesis → chlorophyll → plants → carbon dioxide
• Provides richer, more complete context

Real-world analogy:

• Traditional: Search "Paris" → get Paris documents
• Hybrid: Search "Paris" → get Paris + France + Eiffel Tower + Seine River
• Graph connections provide context vectors can't capture!

Constructors

HybridGraphRetriever(KnowledgeGraph<T>, IDocumentStore<T>)

Initializes a new instance of the HybridGraphRetriever<T> class.

public HybridGraphRetriever(KnowledgeGraph<T> graph, IDocumentStore<T> documentStore)

Parameters

graph KnowledgeGraph<T>

The knowledge graph containing entity relationships.

documentStore IDocumentStore<T>

The document store for similarity search.

Methods

Retrieve(Vector<T>, int, int, int)

Retrieves relevant nodes using hybrid vector + graph approach.

public List<RetrievalResult<T>> Retrieve(Vector<T> queryEmbedding, int topK = 5, int expansionDepth = 1, int maxResults = 10)

Parameters

queryEmbedding Vector<T>

The query embedding vector.

topK int

Number of initial candidates to retrieve via vector search.

expansionDepth int

How many hops to traverse in the graph (0 = no expansion).

maxResults int

Maximum total results to return after expansion.

Returns

List<RetrievalResult<T>>

List of retrieved nodes with their relevance scores.

RetrieveAsync(Vector<T>, int, int, int)

Retrieves relevant nodes asynchronously using hybrid approach.

public Task<List<RetrievalResult<T>>> RetrieveAsync(Vector<T> queryEmbedding, int topK = 5, int expansionDepth = 1, int maxResults = 10)

Parameters

queryEmbedding Vector<T>

topK int

expansionDepth int

maxResults int

Returns

Task<List<RetrievalResult<T>>>

RetrieveWithRelationships(Vector<T>, int, Dictionary<string, double>?, int)

Retrieves nodes with relationship-aware scoring.

public List<RetrievalResult<T>> RetrieveWithRelationships(Vector<T> queryEmbedding, int topK = 5, Dictionary<string, double>? relationshipWeights = null, int maxResults = 10)

Parameters

queryEmbedding Vector<T>

The query embedding vector.

topK int

Number of initial candidates.

relationshipWeights Dictionary<string, double>

Weights for different relationship types.

maxResults int

Maximum results to return.

Returns

List<RetrievalResult<T>>

List of retrieved nodes with relationship-aware scores.