dev.langchain4j.store.embedding

Interface EmbeddingStore<Embedded>

Type Parameters:

Embedded - The class of the object that has been embedded. Typically, this is TextSegment.

public interface EmbeddingStore<Embedded>

Represents a store for embeddings, also known as a vector database.

Method Summary

Modifier and Type	Method and Description
String	add(Embedding embedding) Adds a given embedding to the store.
String	add(Embedding embedding, Embedded embedded) Adds a given embedding and the corresponding content that has been embedded to the store.
void	add(String id, Embedding embedding) Adds a given embedding to the store.
List<String>	addAll(List<Embedding> embeddings) Adds multiple embeddings to the store.
List<String>	addAll(List<Embedding> embeddings, List<Embedded> embedded) Adds multiple embeddings and their corresponding contents that have been embedded to the store.
default List<EmbeddingMatch<Embedded>>	findRelevant(Embedding referenceEmbedding, int maxResults) Finds the most relevant (closest in space) embeddings to the provided reference embedding.
List<EmbeddingMatch<Embedded>>	findRelevant(Embedding referenceEmbedding, int maxResults, double minScore) Finds the most relevant (closest in space) embeddings to the provided reference embedding.

Method Detail

add

String add(Embedding embedding)

Adds a given embedding to the store.

Parameters:

embedding - The embedding to be added to the store.

Returns:

The auto-generated ID associated with the added embedding.

add

void add(String id,
         Embedding embedding)

Adds a given embedding to the store.

Parameters:

id - The unique identifier for the embedding to be added.

embedding - The embedding to be added to the store.

add

String add(Embedding embedding,
           Embedded embedded)

Adds a given embedding and the corresponding content that has been embedded to the store.

Parameters:

embedding - The embedding to be added to the store.

embedded - Original content that was embedded.

Returns:

The auto-generated ID associated with the added embedding.

addAll

List<String> addAll(List<Embedding> embeddings)

Adds multiple embeddings to the store.

Parameters:

embeddings - A list of embeddings to be added to the store.

Returns:

A list of auto-generated IDs associated with the added embeddings.

addAll

List<String> addAll(List<Embedding> embeddings,
                    List<Embedded> embedded)

Adds multiple embeddings and their corresponding contents that have been embedded to the store.

Parameters:

embeddings - A list of embeddings to be added to the store.

embedded - A list of original contents that were embedded.

Returns:

A list of auto-generated IDs associated with the added embeddings.

findRelevant

default List<EmbeddingMatch<Embedded>> findRelevant(Embedding referenceEmbedding,
                                                    int maxResults)

Finds the most relevant (closest in space) embeddings to the provided reference embedding. By default, minScore is set to 0, which means that the results may include embeddings with low relevance.

Parameters:

referenceEmbedding - The embedding used as a reference. Returned embeddings should be relevant (closest) to this one.

maxResults - The maximum number of embeddings to be returned.

Returns:

A list of embedding matches. Each embedding match includes a relevance score (derivative of cosine distance), ranging from 0 (not relevant) to 1 (highly relevant).

findRelevant

List<EmbeddingMatch<Embedded>> findRelevant(Embedding referenceEmbedding,
                                            int maxResults,
                                            double minScore)

Finds the most relevant (closest in space) embeddings to the provided reference embedding.

Parameters:

referenceEmbedding - The embedding used as a reference. Returned embeddings should be relevant (closest) to this one.

maxResults - The maximum number of embeddings to be returned.

minScore - The minimum relevance score, ranging from 0 to 1 (inclusive). Only embeddings with a score of this value or higher will be returned.

Returns:

A list of embedding matches. Each embedding match includes a relevance score (derivative of cosine distance), ranging from 0 (not relevant) to 1 (highly relevant).