A new open-source library, swift-huggingface, has been released, offering Swift developers a native client to interact with Hugging Face's APIs. The library provides complete coverage of the Hugging Face Inference API and Hub API, including text generation, image classification, and model management.
Key features include:
- Inference API: Supports text, image, audio, and multimodal models with streaming, async/await concurrency, and cancellation.
- Hub API: Manage models, datasets, Spaces, and repositories with full CRUD operations.
- Swift-first design: Uses Swift's type system with Codable models and async/await for modern concurrency.
- Extensible: Open-source under MIT license, with modular architecture for adding custom endpoints.
The library aims to integrate with iOS, macOS, watchOS, tvOS, and server-side Swift projects, reducing boilerplate code for AI app development. Developers can install it via Swift Package Manager.