Package dev.langchain4j.store.embedding.vespa

Class VespaEmbeddingStore

java.lang.Object

dev.langchain4j.store.embedding.vespa.VespaEmbeddingStore

All Implemented Interfaces:

EmbeddingStore<TextSegment>

public class VespaEmbeddingStore
extends Object
implements EmbeddingStore<TextSegment>

Represents the Vespa - search engine and vector database. Does not support storing Metadata yet. Example server configuration contains cosine similarity search rank profile, of course other Vespa neighbor search methods are supported too. Read more here.

Constructor Summary

Constructors

Constructor	Description
VespaEmbeddingStore(String url, String keyPath, String certPath, Duration timeout, String namespace, String documentType, String rankProfile, Integer targetHits, Boolean avoidDups)	Creates a new VespaEmbeddingStore instance.

Method Summary

Modifier and Type	Method	Description
String	add(Embedding embedding)
String	add(Embedding embedding, TextSegment textSegment)
void	add(String id, Embedding embedding)	Adds a new embedding with provided ID to the store.
List<String>	addAll(List<Embedding> embeddings)
List<String>	addAll(List<Embedding> embeddings, List<TextSegment> embedded)
List<EmbeddingMatch<TextSegment>>	findRelevant(Embedding referenceEmbedding, int maxResults, double minScore)	The score inside EmbeddingMatch is Vespa relevance according to provided rank profile.

Methods inherited from class java.lang.Object

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

Methods inherited from interface dev.langchain4j.store.embedding.EmbeddingStore

findRelevant, findRelevant, findRelevant, search

Constructor Details

VespaEmbeddingStore

public VespaEmbeddingStore(String url,
 String keyPath,
 String certPath,
 Duration timeout,
 String namespace,
 String documentType,
 String rankProfile,
 Integer targetHits,
 Boolean avoidDups)

Creates a new VespaEmbeddingStore instance.

Parameters:

url - server url, local or cloud one. The latter you can find under Endpoint of your Vespa application, e.g. https://alexey-heezer.langchain4j.mytenant346.aws-us-east-1c.dev.z.vespa-app.cloud/

keyPath - local path to the SSL private key file in PEM format. Read docs for details.

certPath - local path to the SSL certificate file in PEM format. Read docs for details.

timeout - for Vespa Java client in java.time.Duration format.

namespace - required for document ID generation, find more details here.

documentType - document type, used for document ID generation, find more details here and data querying

rankProfile - rank profile from your .sd schema. Provided example schema configures cosine similarity match

targetHits - sets the number of hits (10 is default) exposed to the real Vespa's first-phase ranking function per content node, find more details here.

avoidDups - if true (default), then VespaEmbeddingStore will generate a hashed ID based on provided text segment, which avoids duplicated entries in DB. If false, then random ID will be generated.

Method Details

add

public String add(Embedding embedding)

Specified by:

add in interface EmbeddingStore<TextSegment>

add

public void add(String id,
 Embedding embedding)

Adds a new embedding with provided ID to the store.

Specified by:

add in interface EmbeddingStore<TextSegment>

Parameters:

id - "user-specified" part of document ID, find more details here

embedding - the embedding to add

add

public String add(Embedding embedding,
 TextSegment textSegment)

Specified by:

add in interface EmbeddingStore<TextSegment>

addAll

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

Specified by:

addAll in interface EmbeddingStore<TextSegment>

addAll

public List<String> addAll(List<Embedding> embeddings,
 List<TextSegment> embedded)

Specified by:

addAll in interface EmbeddingStore<TextSegment>

findRelevant

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

The score inside EmbeddingMatch is Vespa relevance according to provided rank profile.

Specified by:

findRelevant in interface EmbeddingStore<TextSegment>