AzureAISearchOptions

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:101

Embeddings and documents are stored in an Azure AI Search index, a merge or upload approach is used when adding embeddings. When adding multiple embeddings the index is updated by this vector store in batches of 10 documents, very large nodes may result in failure due to the batch byte size being exceeded.

Type Parameters

• T extends R

Properties

userAgent?

optional userAgent: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:102

credential?

optional credential: DefaultAzureCredential | AzureKeyCredential | ManagedIdentityCredential

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:103

endpoint?

optional endpoint: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:107

key?

optional key: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:108

serviceApiVersion?

optional serviceApiVersion: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:109

indexName?

optional indexName: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:110

indexClient?

optional indexClient: SearchIndexClient

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:111

indexManagement?

optional indexManagement: IndexManagement

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:112

searchClient?

optional searchClient: SearchClient<T>

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:113

languageAnalyzer?

optional languageAnalyzer: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:114

compressionType?

optional compressionType: KnownVectorSearchCompressionKind

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:115

embeddingDimensionality?

optional embeddingDimensionality: number

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:116

vectorAlgorithmType?

optional vectorAlgorithmType: KnownVectorSearchAlgorithmKind

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:117

idFieldKey?

optional idFieldKey: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:121

Index field storing the id

chunkFieldKey?

optional chunkFieldKey: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:125

Index field storing the node text

embeddingFieldKey?

optional embeddingFieldKey: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:129

Index field storing the embedding vector

metadataStringFieldKey?

optional metadataStringFieldKey: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:136

Index field storing node metadata as a json string. Schema is arbitrary, to filter on metadata values they must be stored as separate fields in the index, use filterable_metadata_field_keys to specify the metadata values that should be stored in these filterable fields

docIdFieldKey?

optional docIdFieldKey: string

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:140

Index field storing doc_id

hiddenFieldKeys?

optional hiddenFieldKeys: string[]

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:146

List of index fields that should be hidden from the client. This is useful for fields that are not needed for retrieving, but are used for similarity search, like the embedding field.

filterableMetadataFieldKeys?

optional filterableMetadataFieldKeys: FilterableMetadataFieldKeysType

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:147

indexMapping()?

optional indexMapping: (enrichedDoc, metadata) => T

Defined in: providers/storage/azure/src/vectorStore/AzureAISearchVectorStore.ts:163

(Optional) function used to map document fields to the AI search index fields If none is specified a default mapping is provided which uses the field keys. The keys in the enriched document are: ["id", "chunk", "embedding", "metadata"].

The default mapping is:

"id" to idFieldKey
"chunk" to chunkFieldKey
"embedding" to embeddingFieldKey
"metadata" to metadataFieldKey