Bagel JavaScript Client 🥯

Table of Contents

• Installation
• Overview
• Client
• Settings
• Usage
• API Methods
  • Ping API
  • Get API Version
  • Create Asset
    • Create Asset: RAW Type Dataset
    • Create Asset: VECTOR Type Dataset
  • Add Embeddings: 'VECTOR' Type Asset
  • Query Vector Asset
  • Get Asset by ID
  • Get All Assets
  • Upload File to Asset
  • Update Asset
  • Get User Details
  • Create API Key
  • Fine-tuning
  • Get Job by Asset
  • Get Job
  • List Jobs
  • List Model Files
  • Delete Asset

Installation

To install the Bagel JavaScript client, use npm:

npm install bagelml

Overview

The official Bagel API endpoint is api.bageldb.ai. The Bagel JavaScript client provides easy access to the Bagel API from Node.js applications.

Also, install the following dependencies:

1. Axios : npm install axios
2. Node-Fetch: npm install node-fetch
3. Form data: npm install form-data
4. uuid: npm install uuid
5. Buffer: npm install buffer

The full source code with examples is available on GitHub.

Client

The Client class is the main interface to the Bagel API. It requires a Settings object to configure connectivity:

import { Settings, Client } from "bageldb-beta";

// Settings config
const settings = new Settings({
  bagel_api_impl: "rest",
  bagel_server_host: "api.bageldb.ai",
});

const client = new Client(settings);

Settings

The Settings class contains configuration options for the client:

• bagel_api_impl - The Bagel API implementation, usually "rest"
• bagel_server_host - Bagel server hostname
• See Settings source for additional options

Usage

Once you have created a Client instance, you can call API methods as shown in the examples below.

API Methods

Ping API

const pingExample = async () => {
  const response = await client.ping();
  console.log(response);
};

pingExample();

This method pings the API to check connectivity. You will get Pong! as a response, acknowledging that the Bagel API is reachable.

Get API Version

const versionExample = async () => {
  const version = await client.get_version();
  console.log(version);
};

versionExample();

This method retrieves the API version string.

Create Asset

Assets in Bagel serve as powerful containers for large datasets, encapsulating embeddings — high-dimensional vectors that represent various data forms, such as text, images, or audio. These Assets enable efficient similarity searches, which are fundamental to a wide range of applications, from recommendation systems and search engines to data analytics tools.

Create Asset: 'RAW' Type dataset

A raw dataset is an unprocessed collection of data in its original form. This type of dataset typically consists of various types of data such as text, images, audio, or any other form of data that hasn't been transformed, processed, or encoded into a specific format.

const apiKey = "insert-your-api-key";
const payload = {
  dataset_type: "RAW",
  title: "Insert Asset Name",
  category: "Insert Category",
  details: "Insert Details",
  tags: [],
  userId: "Insert Your User ID",
};

const createAsset = async () => {
  const asset = await client.create_asset(payload, apiKey);
  console.log(asset);
};

createAsset();

This method creates a new asset and returns a response indicating "Asset successfully created" along with the asset ID. If the asset already exists, the response will be:

data: {
  error: "ValueError('Asset already exists')";
}

NOTE: Ensure all assets you create are unique to avoid errors.

Create Asset: 'VECTOR' Type Asset

A vector dataset consists of data that has been transformed into vectors, which are numerical representations of the original data. Each vector typically contains a set of numbers (features) that capture the essential characteristics of the data.

Creating a 'VECTOR' type asset is similar to creating a 'RAW' type asset. The only difference is the payload:

const apiKey = "insert-your-api-key";
const payload = {
  dataset_type: "VECTOR",
  title: "Insert Asset Name",
  category: "Insert Category",
  details: "Insert Details",
  tags: [],
  userId: "Insert Your User ID",
  embedding_model: "bagel-text",
  dimensions: 768,
};

const createVectorAsset = async () => {
  const asset = await client.create_asset(payload, apiKey);
  console.log(asset);
};

createVectorAsset();

Add Embeddings: 'VECTOR' Type Asset

const assetId = "insert your asset Id";
const apiKey = "insert your api key";

const payload = {
  metadatas: [{ source: "insert text" }],
  documents: ["Insert text"],
  ids: ["jkfbnjfd-t84urb54hurugb-uuybdiubviwd"], //manually generated by you
};

const addVectorAsset = async () => {
  try {
    console.log("Sending request with payload:", payload);

    const response = await client.add_data_to_asset(assetId, payload, apiKey);

    console.log("Response received:", response);
  } catch (error) {
    console.error("Error embedding data to vector asset:", error);
  }
};

addVectorAsset();

Query Vector Asset

const assetId = "insert your asset Id";
const apiKey = "insert your api key";

const payload = {
  where: {
    // category: 'Cat2',
  },
  where_document: {
    // is_published: true,
  },
  // query_embeddings: [em],
  n_results: 1,
  include: ["metadatas", "documents", "distances"],
  query_texts: ["input query text"],
  padding: false,
};
const query = async () => {
  try {
    console.log("Sending request with payload:", payload);

    const response = await client.query_asset(assetId, payload, apiKey);

    console.log("Response received:", response);
  } catch (error) {
    console.error("Error querryin asset:", error);
  }
};

query();

Get Asset by ID

This method retrieves details for a specific Asset using the generated "Asset ID". An API key is used to ensure security.

const apiKey = "insert-your-api-key";
const assetId = "insert-your-asset-id";

const getAsset = async () => {
  const asset = await client.get_asset_by_Id(assetId, apiKey);
  console.log(asset);
};

getAsset();

Update Asset

Updates data in the Asset

import { Settings, Client } from "bageldb-beta";

// Settings config
const settings = new Settings({
  bagel_api_impl: "rest",
  bagel_server_host: "api.bageldb.ai",
});

const client = new Client(settings);

const assetId = "f4013273-03fa-4d2a-bfb4-d36bda4d5a1c";
const apiKey = "4gB2wJPByf8qnUihAmH8dgbGYsZESEOH";

const payload = {
  price: 200,
  is_published: true,
  is_purchased: true,
  details: "This is for everyone video TV gadget",
  title: "LG Televisoin",
};
const update = async () => {
  try {
    console.log("Sending request with payload:", payload);

    const response = await client.update_asset(assetId, payload, apiKey);

    console.log("Response received:", response);
  } catch (error) {
    console.error("Error update cluster embedding:", error);
  }
};

update();

Get All Assets (For a Specific User)

Retrieves all assets associated with a specific user. An API key is used to ensure security.

const apiKey = "insert-your-api-key";
const userId = "insert-your-user-id";

const getAssets = async () => {
  const assets = await client.get_all_assets(userId, apiKey);
  console.log(assets);
};

getAssets();

Upload File to Asset

This method uploads files to a specific Asset.

const assetId = "insert your asset Id"
const filePath = "./sample_data.csv"
const apiKey = "insert your api key"

const uploadFile = async () => {
  // get version
  const asset = await client.add_file(assetId, filePath, apiKey)
  console.log(asset)
}

uploadFile()

Update Asset

Updates data in the Asset

const apiKey = "insert-your-api-key";
const assetId = "insert-your-asset-id";

const payload = {
  price: 200,
  is_published: true,
  is_purchased: true,
  details: "insert text",
  title: "insert text",
};
const update = async () => {
  try {
    console.log("Sending request with payload:", payload);

    const response = await client.update_asset(assetId, payload, apiKey);

    console.log("Response received:", response);
  } catch (error) {
    console.error("Error update cluster embedding:", error);
  }
};

update();

Get User Details

Retrieve details of a specific user from BagelDB using their user ID.

const apiKey = "insert your api key";
const userId = "insert your user id"; // Replace with an actual user ID

const getUserDetails = async () => {
  try {
    const userDetails = await client.get_user_details(userId, apiKey);
    console.log(userDetails);
  } catch (error) {
    console.error("Error retrieving user details:", error);
  }
};

getUserDetails();

Create API Key

Create a new API key for a specified user.

const userId = "insert your user id";

const createApiKey = async (userId) => {
  try {
    const apiKeyDetail = await client.create_api_key("api-key1", userId);
    console.log(apiKeyDetail);
  } catch (error) {
    console.error("Error creating API key:", error);
  }
};

createApiKey(userId);

Finetuning

Fine-tune a model using a specific dataset and configuration.

const apiKey = "insert your api key";

const payload = {
  dataset_type: "MODEL",
  title: "insert title", // 
  category: "insert category", 
  details: "insert details", //choose detail
  tags: [],
  userId: "insert user id", //your user id 
  fine_tune_payload: {
    asset_id: "insert RAW asset id", // make sure to upload a .txt file to the raw asset after creating
    model_name: "insert model name ", // Same name as the title 
    base_model: "insert base model id", // asset id of purchased model from marketplace 
    file_name: "nameoffile.txt", // file name in RAW asset 
    userId: "insert your user id", // your user id 
  },
};

// Function to initiate fine-tuning
const testFineTune = async () => {
  try {
    console.log("Sending request with payload:", payload);
    const response = await client.fine_tune(payload, apiKey);
    console.log('Fine tune response:', response);
  } catch (error) {
    console.error('Error during fine tuning:', error);
  }
};

testFineTune();

Get Job by asset

Retrieve job details associated with a specific asset ID.

const apiKey = "insert your api key";
const asset_id = "insert your asset id"; // Replace with actual asset ID

const testGetJobByAsset = async () => {
  try {
    const response = await client.get_job_by_asset(asset_id, apiKey);
    console.log('Get job by asset response:', response);
  } catch (error) {
    console.error('Error getting job by asset:', error.message);
    console.error('Error details:', error);
  }
};

testGetJobByAsset();

Get Job

Retrieve details of a specific job by its ID.

const apiKey = "insert your api key";
const job_id = "insert your job id"; // Replace with actual job ID

const testGetJob = async () => {
  try {
    const response = await client.get_job(job_id, apiKey);
    console.log('Get job response:', response);
  } catch (error) {
    console.error('Error getting job:', error.message);
    console.error('Error details:', error);
  }
};

testGetJob();

List Jobs

List all jobs associated with a specific user.

const apiKey = "insert your api key";
const user_id = "insert your user id"; // Replace with actual user ID

const testListJobs = async () => {
  try {
    const response = await client.list_jobs(user_id, apiKey);
    console.log('List jobs response:', response);
  } catch (error) {
    console.error('Error listing jobs:', error.message);
    console.error('Error details:', error);
  }
};

testListJobs();

List model files

List all files associated with a specific model asset.

const apiKey = "insert your api key";
const asset_id = "insert your asset id"; // Replace with actual asset ID

const testListModelFiles = async () => {
  try {
    const response = await client.list_model_files(asset_id, apiKey);
    console.log('List model files response:', response);
  } catch (error) {
    console.error('Error listing model files:', error.message);
    console.error('Error details:', error);
  }
};

testListModelFiles();

Delete Asset

This method deletes a specific Asset.

const apiKey = "insert your api key";
const assetId = "insert your asset id"; // Replace with actual asset ID

const deleteAsset = async () => {
  // get version
  const asset = await client.delete_asset(assetId, apiKey);
  console.log(asset);
};
deleteAsset();

This documentation provides examples of using Bagels API methods for user management, job handling, model management, and finetuning. For additional support, please contact [email protected] 🥯.

All changes can be viewed on your console.