A step-by-step guide to building a React Native chat app that runs a language model entirely on-device using the QVAC SDK and Expo. Covers QVAC installation, model downloading with progress tracking, GPU-accelerated model loading, and real-time streaming token generation — all without any cloud dependency. Uses Llama 3.2 1B in GGUF format and requires a physical device for testing since emulators are unsupported.
Table of contents
Table of ContentsPrerequisitesWhat is QVAC?Environment SetupModel ManagementComplete ImplementationConclusion331 Impressions