Skip to content
Frame

Using the Frame SDK

The Frame SDK provides a simple Typescript-based interface for interacting with your deployment. At its core, the SDK revolves around the concept of a document.

What is a “Document”?

In Frame, a “document” represents a unit of data that gets embedded and stored in your vector database. Because Frame is multimodal, a document can be:

  • A piece of text (e.g., a paragraph, product description)
  • An image (e.g., a photo or screenshot)
  • A combination of both text and image (e.g., an image with a caption)

We developed Frame to be flexible, allowing developers to adapt the platform to their specific embedding needs.

Installation

Terminal window
npm install frame-v1

Usage

import { Client } from "frame-v1`";


const frame = new Client({
  apiKey: "<YOUR_API_KEY>",
  baseURL: "https://<YOUR_API_ENDPOINT>.com",
});


// Example usage
frame.getDocuments().then((response) => {
  console.log(response);
});

getDocuments

Gets a list of documents with optional arguments to paginate results.

Arguments:

  • limit number (optional)
  • offset number (optional)

By default, limit is 20 and offset is 0.

Example:

frame.getDocuments();
frame.getDocuments({ limit: 10, offset: 15 });

getDocumentById

Find a specific document by its ID.

Arguments:

  • id number (required)

Example

frame.getDocumentById(3);

createDocuments

Add new documents. A document may optionally omit a URL or a description, but at least one of either must exist.

Arguments:

  • documents array (required)

Example

const documents = [
  {
    url: "https://media.newyorker.com/photos/59095c501c7a8e33fb38c107/master/pass/Monkey-Selfie-DailyShouts.jpg",
    description: "foo",
    metadata: {
      title: "bar",
      altText: "baz",
    },
  },
  {
    url: "https://media.newyorker.com/photos/59095bb86552fa0be682d9d0/master/pass/Monkey-Selfie.jpg",
  },
];


frame.createDocuments(documents);

deleteDocumentById

Delete a specific document by its ID.

Arguments:

  • id number (required)

Example

frame.deleteDocumentById(1);

searchDocuments

Search similar documents based on input, with optional arguments for topK (number of results) and similarity threshold (0.0 - 1.0).

By default, topK is 10 and threshold is 0.

Arguments:

  • url string (at least one of url or description must exist)
  • description string (at least one of url or description must exist)
  • threshold number (optional)
  • topK number (optional)

Example

const query = {
  url: "https://example.com/monalisa.jpeg",
  description: "An image of a sitting woman",
  threshold: 0.3,
  topK: 5,
};


frame.searchDocuments(query);

getRecommendations

Get similar documents based on a document’s ID.

By default, topK is 10 and threshold is 0.

Arguments:

  • id number (required)
  • topK number (optional)
  • threshold number (optional)

Example

frame.getRecommendations(1, { topK: 5, threshold: 0.3 });