// Programatically interact with the Hub

await createRepo({
  repo: {type: "model", name: "my-user/nlp-model"},
  credentials: {accessToken: HF_TOKEN}
});

await uploadFile({
  repo: "my-user/nlp-model",
  credentials: {accessToken: HF_TOKEN},
  // Can work with native File in browsers
  file: {
    path: "pytorch_model.bin",
    content: new Blob(...) 
  }
});

// Use hosted inference

await inference.translation({
  model: 't5-base',
  inputs: 'My name is Wolfgang and I live in Berlin'
})

await inference.textToImage({
  model: 'stabilityai/stable-diffusion-2',
  inputs: 'award winning high resolution photo of a giant tortoise/((ladybird)) hybrid, [trending on artstation]',
  parameters: {
    negative_prompt: 'blurry',
  }
})

// and much more…

Hugging Face JS libraries

This is a collection of JS libraries to interact with the Hugging Face API, with TS types included.

• @huggingface/inference: Use Inference Endpoints (dedicated) and Inference API (serverless) to make calls to 100,000+ Machine Learning models
• @huggingface/hub: Interact with huggingface.co to create or delete repos and commit / download files
• @huggingface/agents: Interact with HF models through a natural language interface

We use modern features to avoid polyfills and dependencies, so the libraries will only work on modern browsers / Node.js >= 18 / Bun / Deno.

The libraries are still very young, please help us by opening issues!

Installation

From NPM

To install via NPM, you can download the libraries as needed:

npm install @huggingface/inference
npm install @huggingface/hub
npm install @huggingface/agents

Then import the libraries in your code:

import { HfInference } from "@huggingface/inference";
import { HfAgent } from "@huggingface/agents";
import { createRepo, commit, deleteRepo, listFiles } from "@huggingface/hub";
import type { RepoId, Credentials } from "@huggingface/hub";

From CDN or Static hosting

You can run our packages with vanilla JS, without any bundler, by using a CDN or static hosting. Using ES modules, i.e. <script type="module">, you can import the libraries in your code:

<script type="module">
    import { HfInference } from 'https://cdn.jsdelivr.net/npm/@huggingface/[email protected]/+esm';
    import { createRepo, commit, deleteRepo, listFiles } from "https://cdn.jsdelivr.net/npm/@huggingface/[email protected]/+esm";
</script>

Deno

// esm.sh
import { HfInference } from "https://esm.sh/@huggingface/inference"
import { HfAgent } from "https://esm.sh/@huggingface/agents";

import { createRepo, commit, deleteRepo, listFiles } from "https://esm.sh/@huggingface/hub"
// or npm:
import { HfInference } from "npm:@huggingface/inference"
import { HfAgent } from "npm:@huggingface/agents";

import { createRepo, commit, deleteRepo, listFiles } from "npm:@huggingface/hub"

Usage examples

Get your HF access token in your account settings.

@huggingface/inference examples

import { HfInference } from "@huggingface/inference";

const HF_TOKEN = "hf_...";

const inference = new HfInference(HF_TOKEN);

// You can also omit "model" to use the recommended model for the task
await inference.translation({
  model: 't5-base',
  inputs: 'My name is Wolfgang and I live in Amsterdam'
})

await inference.textToImage({
  model: 'stabilityai/stable-diffusion-2',
  inputs: 'award winning high resolution photo of a giant tortoise/((ladybird)) hybrid, [trending on artstation]',
  parameters: {
    negative_prompt: 'blurry',
  }
})

await inference.imageToText({
  data: await (await fetch('https://picsum.photos/300/300')).blob(),
  model: 'nlpconnect/vit-gpt2-image-captioning',  
})

// Using your own dedicated inference endpoint: https://hf.co/docs/inference-endpoints/
const gpt2 = inference.endpoint('https://xyz.eu-west-1.aws.endpoints.huggingface.cloud/gpt2');
const { generated_text } = await gpt2.textGeneration({inputs: 'The answer to the universe is'});

@huggingface/hub examples

import { createRepo, uploadFile, deleteFiles } from "@huggingface/hub";

const HF_TOKEN = "hf_...";

await createRepo({
  repo: "my-user/nlp-model", // or {type: "model", name: "my-user/nlp-test"},
  credentials: {accessToken: HF_TOKEN}
});

await uploadFile({
  repo: "my-user/nlp-model",
  credentials: {accessToken: HF_TOKEN},
  // Can work with native File in browsers
  file: {
    path: "pytorch_model.bin",
    content: new Blob(...) 
  }
});

await deleteFiles({
  repo: {type: "space", name: "my-user/my-space"}, // or "spaces/my-user/my-space"
  credentials: {accessToken: HF_TOKEN},
  paths: ["README.md", ".gitattributes"]
});

@huggingface/agents example

import {HfAgent, LLMFromHub, defaultTools} from '@huggingface/agents';

const HF_TOKEN = "hf_...";

const agent = new HfAgent(
  HF_TOKEN,
  LLMFromHub(HF_TOKEN),
  [...defaultTools]
);


// you can generate the code, inspect it and then run it
const code = await agent.generateCode("Draw a picture of a cat wearing a top hat. Then caption the picture and read it out loud.");
console.log(code);
const messages = await agent.evaluateCode(code)
console.log(messages); // contains the data

// or you can run the code directly, however you can't check that the code is safe to execute this way, use at your own risk.
const messages = await agent.run("Draw a picture of a cat wearing a top hat. Then caption the picture and read it out loud.")
console.log(messages); 

There are more features of course, check each library's README!

Formatting & testing

sudo corepack enable
pnpm install

pnpm -r format:check
pnpm -r lint:check
pnpm -r test

Building

pnpm -r build

This will generate ESM and CJS javascript files in packages/*/dist, eg packages/inference/dist/index.mjs.