Class Typesense

constructor

• new Typesense(embeddings, config): Typesense

• Parameters

  • embeddings: EmbeddingsInterface

  • config: TypesenseConfig

  Returns Typesense

FilterType

addDocuments

• addDocuments(documents): Promise<void>

• Add documents to the vector store. Will be updated if in the metadata there is a document with the same id if is using the default import function. Metadata will be added in the columns of the schema based on metadataColumnNames.

  Parameters

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

    Documents to add.

  Returns Promise<void>

addVectors

• addVectors(vectors, documents): Promise<void>

• Adds vectors to the vector store.

  Parameters

  • vectors: number[][]

    Vectors to add.

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

    Documents associated with the vectors.

  Returns Promise<void>

deleteDocuments

• deleteDocuments(documentIds): Promise<void>

• Delete documents from the vector store.

  Parameters

  • documentIds: string[]

    ids of the documents to delete

  Returns Promise<void>

similaritySearchVectorWithScore

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

• Search for similar documents with their similarity score.

  Parameters

  • vectorPrompt: number[]

    vector to search for

  • Optional k: number

    amount of results to return

  • Optional filter: Partial<MultiSearchRequestSchema>

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

  similar documents with their similarity score

Static fromDocuments

• fromDocuments(docs, embeddings, config): Promise<Typesense>

• Create a vector store from documents.

  Parameters

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

    documents

  • embeddings: EmbeddingsInterface

    embeddings

  • config: TypesenseConfig

    Typesense configuration

  Returns Promise<Typesense>

  Typesense vector store

  Warning

  You can omit this method, and only use the constructor and addDocuments.

Static fromTexts

• fromTexts(texts, metadatas, embeddings, config): Promise<Typesense>

• Create a vector store from texts.

  Parameters

  • texts: string[]

  • metadatas: object[]

  • embeddings: EmbeddingsInterface

  • config: TypesenseConfig

  Returns Promise<Typesense>

  Typesense vector store