Class that provides a wrapper around the OpenSearch service for vector search. It provides methods for adding documents and vectors to the OpenSearch index, searching for similar vectors, and managing the OpenSearch index.

Hierarchy

  • VectorStore
    • OpenSearchVectorStore

Constructors

Properties

FilterType: OpenSearchFilter

Methods

  • Method to add documents to the OpenSearch index. It first converts the documents to vectors using the embeddings, then adds the vectors to the index.

    Parameters

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

      The documents to be added to the OpenSearch index.

    Returns Promise<void>

    Promise resolving to void.

  • Method to add vectors to the OpenSearch index. It ensures the index exists, then adds the vectors and associated documents to the index.

    Parameters

    • vectors: number[][]

      The vectors to be added to the OpenSearch index.

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

      The documents associated with the vectors.

    • Optional options: {
          ids?: string[];
      }

      Optional parameter that can contain the IDs for the documents.

      • Optional ids?: string[]

    Returns Promise<void>

    Promise resolving to void.

  • Builds metadata terms for OpenSearch queries.

    This function takes a filter object and constructs an array of query terms compatible with OpenSearch 2.x. It supports a variety of query types including term, terms, terms_set, ids, range, prefix, exists, fuzzy, wildcard, and regexp. Reference: https://opensearch.org/docs/latest/query-dsl/term/index/

    Parameters

    • filter: undefined | OpenSearchFilter

      The filter object used to construct query terms. Each key represents a field, and the value specifies the type of query and its parameters.

    Returns object

    An array of OpenSearch query terms.

    Example

    // Example filter:
    const filter = {
    status: { "exists": true },
    age: { "gte": 30, "lte": 40 },
    tags: ["tag1", "tag2"],
    description: { "wildcard": "*test*" },

    };

    // Resulting query terms:
    const queryTerms = buildMetadataTerms(filter);
    // queryTerms would be an array of OpenSearch query objects.
  • Method to perform a similarity search on the OpenSearch index using a query vector. It returns the k most similar documents and their scores.

    Parameters

    • query: number[]

      The query vector.

    • k: number

      The number of similar documents to return.

    • Optional filter: OpenSearchFilter

      Optional filter for the OpenSearch query.

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

    Promise resolving to an array of tuples, each containing a Document and its score.

  • Static method to create a new OpenSearchVectorStore from an array of Documents, embeddings, and OpenSearch client arguments.

    Parameters

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

      The documents to be added to the OpenSearch index.

    • embeddings: EmbeddingsInterface

      The embeddings used to convert the documents into vectors.

    • dbConfig: OpenSearchClientArgs

      The OpenSearch client arguments.

    Returns Promise<OpenSearchVectorStore>

    Promise resolving to a new instance of OpenSearchVectorStore.

  • Static method to create a new OpenSearchVectorStore from an array of texts, their metadata, embeddings, and OpenSearch client arguments.

    Parameters

    • texts: string[]

      The texts to be converted into documents and added to the OpenSearch index.

    • metadatas: object | object[]

      The metadata associated with the texts. Can be an array of objects or a single object.

    • embeddings: EmbeddingsInterface

      The embeddings used to convert the texts into vectors.

    • args: OpenSearchClientArgs

      The OpenSearch client arguments.

    Returns Promise<OpenSearchVectorStore>

    Promise resolving to a new instance of OpenSearchVectorStore.

Generated using TypeDoc