Hugging Face JS libraries
A collection of TypeScript libraries for working with the Hugging Face API and Hub: managing repositories, calling more than 100,000 hosted inference models, an MCP client, parsers for the GGUF, DDUF and LeRobot formats, and a small library for building tool-using AI agents.
Hugging Face JS libraries is a set of separate TypeScript packages for interacting with the Hugging Face platform. One package creates, deletes and commits to or downloads files from repositories on huggingface.co; another calls more than 100,000 machine learning models through Hugging Face's serverless inference providers or dedicated inference endpoints. A Model Context Protocol client and a small agent library are built on top of that inference client.
Other packages parse the GGUF and DDUF model file formats, parse LeRobot robotics datasets over ranged HTTP requests, define the Hub's core primitives such as pipeline tasks and model libraries, implement a minimal Jinja templating engine for chat templates, help maintain Ollama compatibility with Hub models, and provide a tiny, model-agnostic library for building tool-using AI agents.
The libraries target modern JavaScript runtimes, including current browsers, Node.js 18 and above, Bun and Deno, and install individually through a package manager or directly from a CDN as ES modules.