QVAC

QVAC is an open-source, local-first AI software development kit developed by Tether for running machine learning models directly on personal devices without cloud services or API keys. It supports Linux, macOS, Windows, Android, and iOS through JavaScript, TypeScript, and Python interfaces. Developers can run language models, generate embeddings, and build retrieval-augmented generation workflows. Its multimodal capabilities include image, video, music, and speech generation, transcription, translation, and OCR. The SDK also supports LoRA fine-tuning and local inference using GGUF models. An OpenAI-compatible HTTP server connects local models to AI coding assistants and other applications. Peer-to-peer model distribution allows devices to exchange model files without relying on centralized hosting.