Class RocksetStore

constructor

• new RocksetStore(embeddings, args): RocksetStore

• Constructs a new RocksetStore

  Parameters

  • embeddings: Embeddings

    Object used to embed queries and page content

  • args: RocksetLibArgs

  Returns RocksetStore

FilterType

client

collectionName

embeddingKey

textKey

workspaceName

Optional filter

similarityMetric

• get similarityMetric(): SimilarityMetric

• Returns SimilarityMetric

• set similarityMetric(metric): void

• Parameters

  • metric: SimilarityMetric

  Returns void

addDocuments

• addDocuments(documents): Promise<undefined | string[]>

• Embeds and adds Documents to the store.

  Parameters

  • documents: Document<Record<string, any>>[]

    The documents to store

  Returns Promise<undefined | string[]>

  The _id's of the documents added

addVectors

• addVectors(vectors, documents): Promise<undefined | string[]>

• Adds vectors to the store given their corresponding Documents

  Parameters

  • vectors: number[][]

    The vectors to store

  • documents: Document<Record<string, any>>[]

    The Documents they represent

  Returns Promise<undefined | string[]>

  The _id's of the added documents

delete

• delete(ids): Promise<void>

• Deletes Rockset documements given their _id's

  Parameters

  • ids: string[]

    The IDS to remove documents with

  Returns Promise<void>

destroy

• destroy(waitUntilDeletion?): Promise<void>

• Deletes the collection this RocksetStore uses

  Parameters

  • Optional waitUntilDeletion: boolean

    Whether to sleep until the collection is ready to be queried

  Returns Promise<void>

similaritySearchVectorWithScore

• similaritySearchVectorWithScore(query, k, filter?): Promise<[Document<Record<string, any>>, number][]>

• Gets the most relevant documents to a query along with their similarity score. The returned documents are ordered by similarity (most similar at the first index)

  Parameters

  • query: number[]

    The embedded query to search the store by

  • k: number

    The number of documents to retreive

  • Optional filter: string

    The SQL WHERE clause to filter by

  Returns Promise<[Document<Record<string, any>>, number][]>

Static fromDocuments

• fromDocuments(docs, embeddings, dbConfig): Promise<RocksetStore>

• Constructs, adds docs to, and returns a RocksetStore object

  Parameters

  • docs: Document<Record<string, any>>[]

    The Documents to store

  • embeddings: Embeddings

    The object used to embed queries and page content

  • dbConfig: RocksetLibArgs

    The options to be passed into the RocksetStore constructor

  Returns Promise<RocksetStore>

Static fromTexts

• fromTexts(texts, metadatas, embeddings, dbConfig): Promise<RocksetStore>

• Constructs and returns a RocksetStore object given texts to store.

  Parameters

  • texts: string[]

    The texts to store

  • metadatas: object | object[]

    The metadatas that correspond to

  • embeddings: Embeddings

    The object used to embed queries and page content

  • dbConfig: RocksetLibArgs

    The options to be passed into the RocksetStore constructor

  Returns Promise<RocksetStore>

Static withNewCollection

• withNewCollection(embeddings, dbConfig, collectionOptions?): Promise<RocksetStore>

• Creates a new Rockset collection and returns a RocksetStore that uses it

  Parameters

  • embeddings: Embeddings

    Object used to embed queries and page content

  • dbConfig: RocksetLibArgs

    The options to be passed into the RocksetStore constructor

  • Optional collectionOptions: CreateCollectionRequest

    The arguments to sent with the HTTP request when creating the collection. Setting a field mapping that VECTOR_ENFORCEs is recommended when using this function. See https://rockset.com/docs/vector-functions/#vector_enforce

  Returns Promise<RocksetStore>