Developers can now tap QVAC SDK, a universal JavaScript/TypeScript library designed for on‑device AI across desktops, mobiles and servers. Released under the Apache 2.0 license, the kit wraps the cross‑platform QVAC Fabric inference engine and runs on Bare, the lightweight runtime from the Pear ecosystem. Built‑in support targets Node, Bun and React Native (Hermes).

The SDK handles local inference for large language models, OCR, translation, transcription, text‑to‑speech and vision workloads. Model distribution leverages the Holepunch stack, enabling peer‑to‑peer seeding similar to BitTorrent, while a plugin architecture lets developers add new engines without rewriting core code. Documentation is formatted for both humans and AI assistants, speeding prototype cycles.

Current limitations include oversized bundles due to Bare add‑on packaging, a plugin workflow that still feels clunky, and tree‑shaking that requires an explicit CLI step. The team promises tighter integration in future releases and invites community feedback. As it stands, QVAC SDK offers a ready‑made path for building scalable, local‑first AI applications.