Introduction

Open Data Library (ODL) is the JS/TS SDK for Adafel Protocol

Open Data Library provides a comprehensive interface for users to interact with the Adafel Protocol used to operate on the data layer. The ODL is used to call the data collection and aggregation smart contracts and provides the following functionalities:

• Schema Creation
• Data Collection
• Data Aggregation
• ML Computation
• Fetching Analytics

Installation

npm i @adafel/opendatalibrary-js-sdk

TODO: Add detailing

Quickstart

1. Initialize the client instance

import { privateKeyToAccount } from 'viem/accounts';
import { OpenDataLibrary, Category } from "@adafel/opendatalibrary-js-sdk";
const privateKey = '0xabc'; // optional

const client = new OpenDataLibrary({
  account: privateKeyToAccount(privateKey), // optional
});

Note: if the private key is not supplied then sdk will use provider from window.ethereum.

2. Reading Data from CSV You can read data from a CSV file and format it for use with the library:

const csvPath = "./data.csv";
const inputColumns = ["column1", "column2"];
const labelColumn = "label";

const [inputData, labels] = await odl.getDataFromCSV(
  csvPath,
  inputColumns,
  labelColumn
);

3. Adding Data to the Blockchain To add data to the blockchain, use the addData method:

const schemaName = "MySchema";
const columns = ["column1", "column2"];
const category = Category.Gaming;
const data = inputData;

await odl.addData(schemaName, columns, category, data);

4. Training a Model You can train different models using the provided methods. For example, to train a linear regression model:

const modelName = "MyLinearRegressionModel";

await odl.trainLinearRegressionOffChainData(inputData, labels, modelName);

5. Upload data to IPFS Incase the data size is large, one can upload data to IPFS

const csvPath = "<PATH TO CSV>";

const cid = await odl.uploadCidDataToIPFS(csvPath);

6. Train model from data CID With Adafel it is possible to train the model by passing the data CID

// Data CID
// Expected data is MxN matrix of integer data
// Eg: [[1, 2, 3], [4, 5, 6]]
const cid = "Qm...";

const cidInput = getCidBytes(cid); // Cid in bytes format
const train_indices = [0, 1...]; // Indices of training columns
const label_index: 2;
const modelName = "my_model";

await trainLinearRegressionFromCid(cidInput, train_indices, label_index, modelName);

7. Making Predictions To make predictions using a trained model:

const predictions = await odl.predictLinearRegressionOnchainModel(
  modelName,
  inputData
);