Reference

Install the dataset SDK:

npm install --save @relevanceai/dataset

​

Quickstart

Import the dataset SDK and create the client.

import { Client, QueryBuilder } from '@relevanceai/dataset';

const client = new Client({
    project: '',
    api_key: '',
    endpoint: '',
});

Create a dataset object and run insert documents and search.

const dataset = client.dataset('1000-movies');

const movies = [{ title: 'Lord of the Rings: The Fellowship of the Ring', grenre: 'action', budget: 100 }, ...]
await dataset.insertDocuments(movies, [{ model_name: 'text-embedding-ada-002', field: 'title' }]);

const {results} = await dataset.search(QueryBuilder().vector('title_vector_', { query: 'LOTR', model: 'text-embeddings-ada-002' }));

​

Methods

​

createIfNotExist

The createIfNotExist method is used to create a dataset if it doesn’t exist.

createIfNotExist(): Promise<Boolean>

​

recreateIfExists

The recreateIfExists method is used to recreate a dataset if it exists.

recreateIfExists(): Promise<Boolean>

​

insertDocuments

The insertDocuments method is used to insert documents into the dataset. It takes in a list of documents and a list of vectorization options.

insertDocuments(documents: array[object], vectorizationOptions?: array[object], options?: Options): Promise<BulkInsertOutput>

​

search

The search method is used to search the dataset. It takes in a query builder object.

search(queryBuilder: QueryBuilder, options?: Options): Promise<SearchOutput>

​

updateDocuments

The updateDocuments method is used to update documents in the dataset. It takes in a list of documents.

updateDocuments(documents: array[object], options?: Options): Promise<BulkUpdateOutput>

​

updateDocumentsWhere

The updateDocumentsWhere method is used to update documents in the dataset based on a filter. It takes in a filter and a list of documents.

updateDocumentsWhere(filter: QueryBuilder, partialUpdates: array[object], options?: Options): Promise<BulkUpdateOutput>

​

getDocument

The getDocument method is used to get a document from the dataset. It takes in a document id.

getDocument(id: string): Promise<GetDocumentOutput>

​

deleteDocument

The deleteDocument method is used to delete a document from the dataset. It takes in a document id.

deleteDocument(id: string): Promise<DeleteDocumentOutput>

​

deleteDocumentsWhere

The deleteDocumentsWhere method is used to delete documents from the dataset based on a filter. It takes in a filter.

deleteDocumentsWhere(filter: QueryBuilder): Promise<DeleteDocumentsOutput>

​

QueryBuilder

The query builder is used to construct queries for searching the dataset. It is a chainable API that allows you to build complex queries with a simple syntax.

To use it, start with QueryBuilder() and chain it with the following methods to generate the query.

​

Vector

The vector method is used to search for documents based on a vector. It takes in a field name and a vector object.

vector(field: string, options: { query?: string, model?: enum, vector?: array[double] }): QueryBuilder

Example:

QueryBuilder().vector('title_vector_', {
    query: 'LOTR',
    model: 'text-embeddings-ada-002',
});

​

Sort

The sort method is used to sort the results based on a field. It takes in a field name and a sort order.

sort(field: string, direction: enum): QueryBuilder

Example:

QueryBuilder().sort('budget', 'asc');

​

Page

The page method is used to paginate the results. It takes in a page number.

page(page: number): QueryBuilder

Example:

QueryBuilder().page(2);

​

Page size

The pageSize method is used to set the page size. It takes in a page size.

pageSize(pageSize: number): QueryBuilder

Example:

QueryBuilder().pageSize(10);

​

Include fields

The includeFields method is used to include fields in the results. It takes in a list of field names.

includeFields(fields: array[string]): QueryBuilder

Example:

QueryBuilder().includeFields(['title', 'budget']);