Install the package
Requirements
| Environment | Minimum Version |
|---|
| Node.js | >= 20.1.0 |
| Browsers | ES2020+ (Chrome 80+, Firefox 78+, Safari 14+, Edge 80+) |
| TypeScript | >= 5.0 (recommended) |
numpy-ts has zero runtime dependencies. The installed size is the library itself.
CDN usage (browsers)
You can load numpy-ts directly in the browser without a build step:
<script type="module">
import * as np from 'https://esm.sh/numpy-ts';
const a = np.array([[1, 2], [3, 4]]);
console.log(a);
</script>
Choose your entry point
numpy-ts ships three entry points. Pick the one that fits your use case:
| Entry Point | Import Path | Returns | Bundle Size | Use When |
|---|
| Full | numpy-ts | NDArray | ~200-300 KB | You want method chaining and the complete API |
| Core | numpy-ts/core | NDArrayCore | ~10-40 KB | Bundle size matters; you only need specific functions |
| Node | numpy-ts/node | NDArray | ~200-300 KB + fs | You need file I/O (load, save, savez) on Node.js |
Full library (numpy-ts)
The default entry point. Functions return NDArray, which supports method chaining (.add(), .reshape(), .T, etc.).
import * as np from 'numpy-ts';
const a = np.array([1, 2, 3, 4]);
const b = a.add(10).reshape([2, 2]).T; // Method chaining
Tree-shakeable core (numpy-ts/core)
Functions return NDArrayCore (no method chaining). Your bundler (webpack, Vite, esbuild, Rollup) will only include the functions you actually import.
import { array, add, reshape, transpose } from 'numpy-ts/core';
const a = array([1, 2, 3, 4]);
const b = transpose(reshape(add(a, 10), [2, 2]));
Importing just array and add from numpy-ts/core results in a bundle around 10 KB minified. The full entry point is always ~200-300 KB regardless of what you use.
Node.js file I/O (numpy-ts/node)
Everything from numpy-ts plus file system operations for .npy, .npz, and text files.
import * as np from 'numpy-ts/node';
// Load and save .npy files
const arr = await np.loadNpy('data.npy');
await np.saveNpy('output.npy', arr);
// Load and save .npz archives
const { arrays } = await np.loadNpzFile('data.npz');
await np.savez('output.npz', { x: arr, y: arr });
// Text files
const data = await np.loadtxt('data.csv', { delimiter: ',' });
await np.savetxt('output.csv', data, { delimiter: ',' });
The numpy-ts/node entry point uses Node.js built-in fs modules. It will not work in browsers or edge runtimes.
Next steps
