Skip to main content

​
Requirements

Before installing Deepbox, ensure your system meets the following requirements:
Node.js Version: Deepbox requires Node.js >= 24.13.0
Check your Node.js version: 
node --version
If you need to upgrade, download the latest version from nodejs.org.

​
Install Deepbox

Install Deepbox using your preferred package manager: 
npm install deepbox

​
Verify Installation

Create a simple test file to verify your installation:
test.ts
import { tensor, add } from "deepbox/ndarray";

const a = tensor([1, 2, 3]);
const b = tensor([4, 5, 6]);
const c = add(a, b);

console.log(c.toString()); // Output: tensor([5, 7, 9])
Run the file: 
npx tsx test.ts
If you see tensor([5, 7, 9]), congratulations! Deepbox is installed correctly.

​
Import Strategies

Deepbox provides flexible import strategies to suit your needs: Import from specific modules for optimal tree-shaking and smaller bundle sizes: 
import { tensor, add, mul } from "deepbox/ndarray";
import { DataFrame } from "deepbox/dataframe";
import { LinearRegression } from "deepbox/ml";
import { StandardScaler } from "deepbox/preprocess";
import { accuracy, f1Score } from "deepbox/metrics";

​
Namespace Imports

Use namespace imports when you prefer explicit module prefixes: 
import * as db from "deepbox";

const t = db.ndarray.tensor([1, 2, 3]);
const df = new db.dataframe.DataFrame({ col: [1, 2, 3] });
const model = new db.ml.LinearRegression();

​
Available Modules

Deepbox is organized into the following modules:
Types, errors, validation, dtype helpers, and configuration
import type { DType, Shape, Device } from "deepbox/core";
N-D tensors with autograd, broadcasting, 90+ operations, and sparse matrices
import { tensor, add, matmul, parameter } from "deepbox/ndarray";
SVD, QR, LU, Cholesky, eigenvalue decomposition, solvers, and norms
import { svd, qr, solve, inv } from "deepbox/linalg";
DataFrame and Series with 50+ operations and CSV I/O
import { DataFrame, Series } from "deepbox/dataframe";
Descriptive stats, correlations, and hypothesis tests
import { mean, std, ttest, anova } from "deepbox/stats";
40+ ML metrics for classification, regression, and clustering
import { accuracy, f1Score, r2Score } from "deepbox/metrics";
Scalers, encoders, normalizers, and cross-validation splits
import { StandardScaler, trainTestSplit } from "deepbox/preprocess";
Classical ML models (Linear, Ridge, Lasso, Trees, SVM, KNN, Ensembles)
import { LinearRegression, RandomForestClassifier } from "deepbox/ml";
Neural network layers (Linear, Conv, RNN/LSTM/GRU, Attention, Normalization)
import { Sequential, Linear, ReLU, mseLoss } from "deepbox/nn";
Optimizers (SGD, Adam, AdamW, RMSprop) and LR schedulers
import { Adam, CosineAnnealingLR } from "deepbox/optim";
Distributions (uniform, normal, binomial, gamma, beta) and sampling
import { normal, uniform, choice } from "deepbox/random";
Built-in datasets (Iris, Digits, Breast Cancer) and synthetic generators
import { loadIris, makeClassification } from "deepbox/datasets";
SVG/PNG plotting (scatter, line, bar, histogram, heatmap, contour)
import { scatter, plot, hist, heatmap, saveFig } from "deepbox/plot";

​
TypeScript Configuration

For the best development experience, configure your tsconfig.json:
tsconfig.json
{
  "compilerOptions": {
    "target": "ES2022",
    "module": "ESNext",
    "moduleResolution": "bundler",
    "esModuleInterop": true,
    "strict": true,
    "skipLibCheck": true,
    "resolveJsonModule": true,
    "isolatedModules": true,
    "lib": ["ES2022"]
  }
}
Enable strict mode to take full advantage of Deepbox’s type safety.

​
Package Information

Deepbox is published to npm with the following characteristics:
  • Package name: deepbox
  • Current version: 0.2.0
  • License: MIT
  • Bundle format: ESM and CommonJS
  • Type definitions: Included
  • Runtime dependencies: Zero

​
Module Exports

Deepbox uses modern ESM exports with CommonJS fallbacks:
package.json
{
  "main": "./dist/index.cjs",
  "module": "./dist/index.js",
  "types": "./dist/index.d.ts"
}
Every module has dedicated exports for tree-shaking: 
// ESM
import { tensor } from "deepbox/ndarray";

// CommonJS
const { tensor } = require("deepbox/ndarray");

​
Next Steps

Quick Start Guide

Build your first Deepbox application in 5 minutes

Core Concepts

Learn about tensors, autograd, and broadcasting

API Reference

Explore the complete API documentation

Examples

See practical examples and use cases

​
Troubleshooting

If you encounter module resolution errors, ensure:
  1. Your Node.js version is >= 24.13.0
  2. Your tsconfig.json has "moduleResolution": "bundler" or "node16"
  3. You’re using a compatible bundler (Vite, esbuild, Rollup, Webpack 5+)
If you see TypeScript errors:
  1. Ensure you have "strict": true in your tsconfig.json
  2. Install @types/node if you haven’t already
  3. Clear your TypeScript cache: rm -rf node_modules/.cache
For large tensor operations, you may need to increase Node.js memory:
node --max-old-space-size=4096 your-script.js
Need help? Check the GitHub Issues or open a new issue.

Build docs developers (and LLMs) love

Get started for freeTalk to us