Reference

Plugin catalog

Each QVAC capability maps to a plugin. You enable only the ones your app needs, which keeps the bundle small — omit the list entirely and every built-in plugin ships. Custom plugins install as npm packages and are enabled the same way.

qvac.config.json

{
  "plugins": [
    "@qvac/sdk/llamacpp-completion/plugin",
    "@qvac/sdk/onnx-ocr/plugin"
  ]
}

Built-in plugins

9 available

LLM

qvac-fabric-llm.cpp

@qvac/sdk/llamacpp-completion/plugin

The core inference plugin. Runs large language models locally for text generation and chat, and underpins multimodal conversation, retrieval-augmented generation, and LoRA fine-tuning.

Text generationMultimodalRAGFine-tuning

Embeddings

qvac-fabric-llm.cpp

@qvac/sdk/llamacpp-embedding/plugin

Generates vector embeddings for semantic search, clustering, and retrieval. Required alongside the LLM plugin for any retrieval workflow.

Text embeddingsRAG

ASR (Whisper)

qvac-ext-lib-whisper.cpp

@qvac/sdk/whispercpp-transcription/plugin

Automatic speech recognition through a customized Whisper backend. The general-purpose choice for speech-to-text.

Transcription

ASR (Parakeet)

NVIDIA Parakeet

@qvac/sdk/parakeet-transcription/plugin

An alternative speech recognition path built on NVIDIA's Parakeet model. Same capability as the Whisper plugin, different engine and performance profile.

Transcription

BCI transcription

qvac-ext-lib-whisper.cpp

@qvac/sdk/bci-whispercpp-transcription/plugin

Brain–computer interface transcription via a customized Whisper backend. Transcribes neural signals rather than audio. The most speculative plugin in the catalog, and a clear signal of where Tether intends the platform to go.

Transcription

NMT

Fabric LLM / Bergamot

@qvac/sdk/nmtcpp-translation/plugin

Neural machine translation, text to text. Can run through Fabric LLM or through Bergamot, the translation engine originally built for in-browser use.

Translation

TTS

ONNX Runtime

@qvac/sdk/onnx-tts/plugin

Speech synthesis using the Chatterbox and Supertonic neural TTS models through ONNX Runtime.

Text-to-Speech

OCR

ONNX Runtime

@qvac/sdk/onnx-ocr/plugin

Optical character recognition for extracting text from images. The bridge between a photograph of a document and something a model can read.

OCR

Diffusion

qvac-ext-stable-diffusion.cpp

@qvac/sdk/sdcpp-generation/plugin

Text-to-image and image-to-image generation through a customized Stable Diffusion backend.

Image generation

A note on Bare

The plugins array is bundle-time configuration — it decides what gets packed into the worker. Runtime registration is separate.

On Node.js and Expo the SDK spawns a worker that auto-registers the full built-in set, so you never register anything by hand. Bare runs in-process with no spawned worker, so nothing auto-registers. Register the plugins you use before the first SDK call, or you will see WorkerPluginsNotRegisteredError.

import { plugins } from "@qvac/bare-sdk";
import { llmPlugin } from "@qvac/bare-sdk/llamacpp-completion/plugin";

const sdk = plugins([llmPlugin]);

Tether recommends @qvac/bare-sdk for direct Bare usage — a slim distribution built for explicit assembly. The full @qvac/sdk also runs on Bare.

Full documentation lives at docs.qvac.tether.io.