QVAC developer resources

Everything you need to start building on QVAC, in one place. Because QVAC is new, resources are still spread out — this hub pulls the official sources together with our own practical guides. Bookmark it; we keep it current as the ecosystem grows.

Official QVAC resources

Documentation — docs.qvac.tether.io
The source of truth: compatibility matrix, per-platform install instructions, full API reference with code examples for each capability, and architecture docs. Start here for anything authoritative.

The SDK — @qvac/sdk on npm
The main entry point. A type-safe JS/TS package exposing all QVAC capabilities through one interface. Install with npm install @qvac/sdk.

GitHub — github.com/tetherto/qvac
The open-source monorepo (Apache 2.0). Components live under /packages — the SDK, libraries, CLI, and addons. Also where you'll find "built with QVAC" badges to add to your own project.

QVAC site — qvac.tether.io
The high-level overview of the initiative, the vision, and the flagship apps.

Key concepts to understand first

QVAC Fabric — the inference engine under the SDK, a fork of llama.cpp, handling text generation and embeddings (plus integrated engines like whisper.cpp for speech and Bergamot for translation).
The model registry — a distributed registry for fetching models, with preconfigured model constants so you can load models without managing files by hand. Supports sharded models and pause/resume downloads.
P2P (Holepunch stack) — peer-to-peer model distribution and delegated inference (offload heavy tasks to a more powerful device).
OpenAI-compatible server — an HTTP server that wraps QVAC and speaks the OpenAI API format, so existing OpenAI-style code can point at local inference.

Runtimes and platforms

The SDK runs on:

Node.js (≥22.17) — servers, CLI tools
Bare — lightweight cross-platform
Expo (≥54) — iOS and Android (physical devices)

Targets iOS, Android, macOS, Windows, and Linux from one codebase. Not the browser (by design).

Capabilities reference

Behind one interface: text completion, embeddings, RAG, transcription (speech-to-text), translation, OCR, vision/multimodal, and more — with LoRA fine-tuning and additional backends in the broader ecosystem. → Full capabilities guide

Try it without coding first

If you want to see QVAC in action before building, download QVAC Workbench (Tether's flagship app) and experiment with local models, RAG, transcription, and delegated inference. It's a good way to build intuition for what the SDK can do. → Workbench guide

localhost/AI's practical guides

We're building on QVAC, and we write up what we learn:

QVAC SDK Explained — the conceptual overview
How to Build a QVAC App — step-by-step orientation
On-Device RAG Over Your Own Notes — a worked pipeline from a shipping app
QVAC Agents Explained — tool-calling and on-device agents
QVAC Local AI Use Cases — where on-device fits

Community

QVAC has a community presence via Tether's channels (including a Keet room linked from the QVAC site) and the GitHub repo for issues and contributions. As an open-source project, contributions back to the ecosystem are welcomed.

We're compiling the ecosystem. If you've shipped — or are building — an app on QVAC, we'd love to hear about it and feature it in Best QVAC Apps.

QVAC is evolving quickly; the official docs are always the most current source. We update this hub as things change.