Trans-N-ai/swama

What it solves

Swama is a high-performance machine learning runtime designed for macOS that allows users to run Large Language Models (LLMs), Vision Language Models (VLMs), and audio models (ASR/TTS) locally on Apple Silicon. It eliminates the need for cloud-based AI by providing a native, optimized environment for inference.

How it works

Built on Apple's MLX framework and written in pure Swift, Swama provides three primary interfaces: a macOS menu bar application for easy access, a command-line interface (CLI) for model management and inference, and an OpenAI-compatible API server. It integrates directly with the HuggingFace Hub to automatically download, cache, and manage models using simplified aliases (e.g., qwen3 instead of full repository URLs).

Who it’s for

It is designed for macOS users with Apple Silicon (M1/M2/M3/M4) who want to run powerful AI models locally for privacy, speed, or development purposes, as well as developers who need an OpenAI-compatible local endpoint for their own applications.

Highlights

• Multimodal Capabilities: Supports text generation, image-to-text (VLM), speech-to-text (ASR), and text-to-speech (TTS).
• OpenAI Compatible API: Provides standard endpoints for chat completions, embeddings, and audio processing, including support for tool calling.
• Smart Model Management: Features automatic downloading and caching of models via friendly aliases.
• Native macOS Integration: Includes a dedicated menu bar app and CLI tools optimized for Apple Silicon.