Qvac

QVAC is Tether's open source toolkit for running AI locally on a device, without a cloud service. It centers on a single SDK, available for JavaScript and Python, that exposes more than ten AI tasks across iOS, Android and desktop through one interface. The listed tasks include text generation, embeddings, retrieval augmented generation over local documents, speech to text, offline translation, image generation, text to speech, OCR, vision and LoRA fine tuning, including on mobile. It is aimed at developers who want to add AI to an app without API keys or per user inference bills, and at privacy minded users who want prompts and files to stay on their own hardware. The vendor describes the stack as an ecosystem of five products for running, training, fine tuning and shipping models on device, and devices can also be linked peer to peer. Tether also publishes its own model family under the QVAC Psy name, including QVAC MedPsy, a set of small medical models intended for edge deployment, which are available on Hugging Face and load through the SDK. The project is licensed under Apache 2.0 and the vendor states there are no subscriptions, per token charges or rate limits, since inference runs on the user's device. Source code is on GitHub.