tetherto/qvac

What it solves

QVAC addresses the challenge of running diverse AI workloads—including text, vision, audio, and generative media—locally on consumer hardware. It removes the need for cloud-based APIs, ensuring privacy and offline functionality across desktop and mobile platforms.

How it works

QVAC provides a unified SDK for JavaScript/TypeScript and Python that handles local model inference. It includes an HTTP server that exposes an OpenAI-compatible API, allowing developers to connect existing AI tools to local models. The system uses a peer-to-peer architecture to distribute models, enabling devices to fetch weights directly from one another without relying on centralized servers.

Who it’s for

It is designed for developers building local-first AI applications, embedded systems, or on-premises tools who want a single interface to manage multiple AI capabilities while maintaining data sovereignty.

Highlights

• Cross-platform support: Runs on Linux, macOS, Windows, Android, and iOS.
• Comprehensive AI suite: Supports text generation, RAG, fine-tuning, image/video generation, speech-to-text, and even robot control (VLA).
• OpenAI-compatible: The built-in HTTP server allows seamless integration with existing AI ecosystems.
• Peer-to-peer distribution: Decentralized model registry and relay system for model sharing and connectivity.
• Developer-friendly: Provides native SDKs for Python and JavaScript/TypeScript with built-in progress tracking and model management.