Kuzco

Kuzco is a Swift package for integrating large language models (LLMs) directly into iOS, macOS, and Mac Catalyst apps. Built on `llama.cpp`, it offers customizable prompts, flexible tuning, and async/await-friendly APIs for on-device AI.

Kuzco is an open-source Swift package that enables developers to integrate and run large language models (LLMs) locally on iOS, macOS, and Mac Catalyst apps. Built on the robust `llama.cpp` framework, it ensures privacy by processing all AI inference on-device without network dependencies.

Key features include support for multiple model architectures like LLaMA, Mistral, and Gemma, optimized for Apple Silicon and Intel Macs. It offers modern Swift concurrency with async/await support, flexible tuning options for model performance, and efficient resource management for mobile constraints.

Ideal for developers prioritizing privacy, speed, and reliability in AI-powered apps, Kuzco simplifies local LLM integration with a clean API and comprehensive error handling. It’s production-ready, memory-efficient, and compatible with iOS 15.0+ and macOS 12.0+.