Package com.microsoft.semantickernel.connectors.memory.azurecognitivesearch

Class AzureCognitiveSearchMemoryStore

java.lang.Object

com.microsoft.semantickernel.connectors.memory.azurecognitivesearch.AzureCognitiveSearchMemoryStore

All Implemented Interfaces:

Buildable, MemoryStore

public class AzureCognitiveSearchMemoryStore
extends Object
implements MemoryStore

Semantic Memory implementation using Azure Cognitive Search. For more information about Azure Cognitive Search {@see https://learn.microsoft.com/azure/search/search-what-is-azure-search}

Nested Class Summary

Nested classes/interfaces inherited from interface com.microsoft.semantickernel.memory.MemoryStore

MemoryStore.Builder<T extends MemoryStore>

Constructor Summary

Constructors

Constructor	Description
AzureCognitiveSearchMemoryStore(com.azure.search.documents.indexes.SearchIndexAsyncClient searchIndexAsyncClient, Function<com.azure.search.documents.SearchDocument,MemoryRecord> memoryRecordMapper)	Create a new instance of memory storage using Azure Cognitive Search.
AzureCognitiveSearchMemoryStore(String endpoint, com.azure.core.credential.TokenCredential credentials)	Create a new instance of memory storage using Azure Cognitive Search.
AzureCognitiveSearchMemoryStore(String endpoint, String apiKey)	Create a new instance of memory storage using Azure Cognitive Search.

Method Summary

Modifier and Type	Method	Description
reactor.core.publisher.Mono<Void>	createCollectionAsync(String collectionName)
reactor.core.publisher.Mono<Void>	deleteCollectionAsync(String collectionName)
reactor.core.publisher.Mono<Boolean>	doesCollectionExistAsync(String collectionName)
reactor.core.publisher.Mono<MemoryRecord>	getAsync(String collectionName, String key, boolean withEmbedding)
reactor.core.publisher.Mono<Collection<MemoryRecord>>	getBatchAsync(String collectionName, Collection<String> keys, boolean withEmbedding)
reactor.core.publisher.Mono<List<String>>	getCollectionsAsync()
reactor.core.publisher.Mono<reactor.util.function.Tuple2<MemoryRecord,Float>>	getNearestMatchAsync(String collectionName, Embedding embedding, float minRelevanceScore, boolean withEmbedding)
reactor.core.publisher.Mono<Collection<reactor.util.function.Tuple2<MemoryRecord,Float>>>	getNearestMatchesAsync(String collectionName, Embedding embedding, int limit, float minRelevanceScore, boolean withEmbedding)
reactor.core.publisher.Mono<Collection<reactor.util.function.Tuple2<MemoryRecord,Float>>>	getNearestMatchesAsync(String collectionName, Embedding embedding, int limit, float minRelevanceScore, boolean withEmbedding, Function<com.azure.search.documents.SearchDocument,MemoryRecord> memoryRecordMapper)	Gets the nearest matches to the Embedding of type Float.
reactor.core.publisher.Mono<Void>	removeAsync(String collectionName, String key)
reactor.core.publisher.Mono<Void>	removeBatchAsync(String collectionName, Collection<String> keys)
reactor.core.publisher.Mono<String>	upsertAsync(String collectionName, MemoryRecord record)
reactor.core.publisher.Mono<Collection<String>>	upsertBatchAsync(String collectionName, Collection<MemoryRecord> records)

Methods inherited from class java.lang.Object

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

Constructor Detail

AzureCognitiveSearchMemoryStore

public AzureCognitiveSearchMemoryStore(@Nonnull
                                       String endpoint,
                                       @Nonnull
                                       String apiKey)

Create a new instance of memory storage using Azure Cognitive Search.

Parameters:

endpoint - Azure Cognitive Search URI, e.g. "https://contoso.search.windows.net"

apiKey - Azure API Key

AzureCognitiveSearchMemoryStore

public AzureCognitiveSearchMemoryStore(@Nonnull
                                       String endpoint,
                                       @Nonnull
                                       com.azure.core.credential.TokenCredential credentials)

Create a new instance of memory storage using Azure Cognitive Search.

Parameters:

endpoint - Azure Cognitive Search URI, e.g. "https://contoso.search.windows.net"

credentials - Azure service credentials

AzureCognitiveSearchMemoryStore

public AzureCognitiveSearchMemoryStore(@Nonnull
                                       com.azure.search.documents.indexes.SearchIndexAsyncClient searchIndexAsyncClient,
                                       @Nullable
                                       Function<com.azure.search.documents.SearchDocument,MemoryRecord> memoryRecordMapper)

Create a new instance of memory storage using Azure Cognitive Search.

Parameters:

searchIndexAsyncClient - the Azure search documents index client to use

Method Detail

createCollectionAsync

public reactor.core.publisher.Mono<Void> createCollectionAsync(@Nonnull
                                                               String collectionName)

Specified by:

createCollectionAsync in interface MemoryStore

getCollectionsAsync

public reactor.core.publisher.Mono<List<String>> getCollectionsAsync()

Specified by:

getCollectionsAsync in interface MemoryStore

doesCollectionExistAsync

public reactor.core.publisher.Mono<Boolean> doesCollectionExistAsync(@Nonnull
                                                                     String collectionName)

Specified by:

doesCollectionExistAsync in interface MemoryStore

deleteCollectionAsync

public reactor.core.publisher.Mono<Void> deleteCollectionAsync(@Nonnull
                                                               String collectionName)

Specified by:

deleteCollectionAsync in interface MemoryStore

upsertAsync

public reactor.core.publisher.Mono<String> upsertAsync(@Nonnull
                                                       String collectionName,
                                                       @Nonnull
                                                       MemoryRecord record)

Specified by:

upsertAsync in interface MemoryStore

upsertBatchAsync

public reactor.core.publisher.Mono<Collection<String>> upsertBatchAsync(@Nonnull
                                                                        String collectionName,
                                                                        Collection<MemoryRecord> records)

Specified by:

upsertBatchAsync in interface MemoryStore

getAsync

public reactor.core.publisher.Mono<MemoryRecord> getAsync(@Nonnull
                                                          String collectionName,
                                                          @Nonnull
                                                          String key,
                                                          boolean withEmbedding)

Specified by:

getAsync in interface MemoryStore

getBatchAsync

public reactor.core.publisher.Mono<Collection<MemoryRecord>> getBatchAsync(@Nonnull
                                                                           String collectionName,
                                                                           @Nonnull
                                                                           Collection<String> keys,
                                                                           boolean withEmbedding)

Specified by:

getBatchAsync in interface MemoryStore

getNearestMatchAsync

public reactor.core.publisher.Mono<reactor.util.function.Tuple2<MemoryRecord,Float>> getNearestMatchAsync(@Nonnull
                                                                                                                String collectionName,
                                                                                                                @Nonnull
                                                                                                                Embedding embedding,
                                                                                                                float minRelevanceScore,
                                                                                                                boolean withEmbedding)

Specified by:

getNearestMatchAsync in interface MemoryStore

getNearestMatchesAsync

public reactor.core.publisher.Mono<Collection<reactor.util.function.Tuple2<MemoryRecord,Float>>> getNearestMatchesAsync(@Nonnull
                                                                                                                              String collectionName,
                                                                                                                              @Nonnull
                                                                                                                              Embedding embedding,
                                                                                                                              int limit,
                                                                                                                              float minRelevanceScore,
                                                                                                                              boolean withEmbedding)

Specified by:

getNearestMatchesAsync in interface MemoryStore

getNearestMatchesAsync

public reactor.core.publisher.Mono<Collection<reactor.util.function.Tuple2<MemoryRecord,Float>>> getNearestMatchesAsync(@Nonnull
                                                                                                                              String collectionName,
                                                                                                                              @Nonnull
                                                                                                                              Embedding embedding,
                                                                                                                              int limit,
                                                                                                                              float minRelevanceScore,
                                                                                                                              boolean withEmbedding,
                                                                                                                              @Nullable
                                                                                                                              Function<com.azure.search.documents.SearchDocument,MemoryRecord> memoryRecordMapper)

Gets the nearest matches to the Embedding of type Float. Does not guarantee that the collection exists.

If a memoryRecordMapper is provided this overrides any mapper provided when constructing this memory store.

Parameters:

collectionName - The name associated with a collection of embeddings.

embedding - The Embedding to compare the collection's embeddings with.

limit - The maximum number of similarity results to return.

minRelevanceScore - The minimum relevance threshold for returned results.

withEmbedding - If true, the embeddings will be returned in the memory records.

memoryRecordMapper - a mapper that controls how to map the search document to a memory record

Returns:

A collection of tuples where item1 is a MemoryRecord and item2 is its similarity score as a Float.

removeAsync

public reactor.core.publisher.Mono<Void> removeAsync(@Nonnull
                                                     String collectionName,
                                                     @Nonnull
                                                     String key)

Specified by:

removeAsync in interface MemoryStore

removeBatchAsync

public reactor.core.publisher.Mono<Void> removeBatchAsync(@Nonnull
                                                          String collectionName,
                                                          @Nonnull
                                                          Collection<String> keys)

Specified by:

removeBatchAsync in interface MemoryStore