On-Device AI

AI that runs entirely in your browser

ALLYX ONE AI tools never send your data to a server. The models are downloaded to your device once and run locally — privately, and even offline.

Your text, images, and documents are processed entirely on your device. Nothing is transmitted to ALLYX ONE servers — there are no API calls, no cloud queues, and no third-party AI services involved.

How it works

Model downloads once

The first time you use an AI tool, your browser downloads the model file and saves it to your browser's local storage. This is a one-time download per model.

AI runs locally

Inference runs inside a Web Worker in your browser using Transformers.js with WebGPU acceleration where available, falling back to WebAssembly automatically.

Results stay on-device

All outputs — summaries, corrected text, generated content — are produced locally and displayed directly to you. Nothing ever leaves your browser.

Hardware requirements

Most AI tools work on everyday hardware. A few larger models benefit from extra RAM or a dedicated GPU.

GPU

4 GB+ VRAM recommended for GPU-accelerated tools. Lower-spec GPUs fall back to CPU automatically — tools still work, just slower.

RAM

8 GB minimum; 16 GB recommended for Power Features (Title Generator, Resume Builder).

Storage

Up to 4 GB of free space required to cache all AI models locally in your browser. Individual models range from ~20 MB to ~2.3 GB.

CPU

Any modern processor (2015 or later). Used automatically as a fallback when GPU acceleration is unavailable.

What's stored in your browser

Model weights (Cache Storage) — The actual AI model files (ONNX format) are stored in your browser's Cache Storage API by Transformers.js after the first download. These are the largest files — between ~20 MB and ~2.3 GB per model.

Download consent record (IndexedDB) — A small flag is recorded in IndexedDB once you consent to a model download. This prevents the download dialog from appearing again on future visits.

Durable storage permission — ALLYX ONE requests durable storage from your browser to prevent model files from being evicted automatically. You can revoke this in your browser's site settings at any time.

None of this storage contains your personal data, documents, or any content you processed with the tools. It is limited to model files and a consent flag.

Manage your browser storage

You can remove all locally cached AI model files at any time. Models will be re-downloaded the next time you use an AI tool.

Clearing models does not affect your work or any other browser data — only the locally cached AI model files are removed.

You can also clear site data manually via your browser's settings — look for "Site settings" or "Cookies and site data" for allyxone.com.

Back to home