
Local AI 101: what your own machine can run
A phone or an 8 GB laptop handles most everyday AI jobs, and video needs 32 GB. How much memory local AI needs, and the command that tells you what yours has.
Read MoreInvoice capture is priced per page and works by uploading your supplier list, your prices, and your bank details. We built a demo (an illustrative example, not a QVAC product) that turns folders of invoices into an accounting table on your own machine: no upload, no per-page bill, and the columns are the ones you decide you need.
Typing invoices into a spreadsheet is the least glamorous job in a small business and one of the most reliably paid-for. The products that do it are good, and they share two properties that are easy to accept one at a time and harder to accept together: they charge per page, and they only work once you have uploaded the invoices.
A folder of supplier invoices is not a neutral pile of paper. It is a map of who you buy from, what you pay them, what your margins probably are, and which bank accounts the money moves between. Handing that to a third party is a real decision, and it is usually made by nobody, in a hurry, at the end of a quarter, because the alternative is typing three hundred lines by hand.
Running the model on your own machine removes the decision. There is no upload, so there is no data-processing agreement to read, no retention policy to trust, and nothing in a bucket after you stop paying. And since the model runs locally, there is nothing to pay: the per-page meter, which is exactly the wrong shape for a shoebox of receipts, simply is not there.
So we built QVAC Invoice Manager: an illustrative example, not a product, that reads a folder of invoices and fills in an accounting table, on-device and for free. It is built on the QVAC SDK, and the point is the pattern, so you can clone it and make it your own. See it in our video below:
Repo: github.com/tetherto/qvac-examples (the qvac-invoice-manager-demo example).
Point it at folders, not files, because a year of expenses is never one flat directory.
Requirement | |
|---|---|
RAM | 8 GB with one model at a time, 16 GB or more comfortable |
Disk | About 2.5 GB for the text model and about 1.6 GB for the vision one. |
GPU | Apple Silicon (Metal) or a Vulkan GPU. CPU works, slower |
OS / runtime | macOS 14+, Windows 10+ or Linux; Node.js 22.17 or newer |
Not sure your machine can handle it? Run npx -y @qvac/cli doctor.
git clone https://github.com/tetherto/qvac-examples
cd qvac-examples/qvac-invoice-manager-demo
npm install
npm startClick Choose folders, point it at a year of expenses, and watch the table fill. Four sample documents ship in demo/ so you can try it without supplying your own. QVAC is open source (Apache 2.0) and free. Docs: docs.qvac.tether.io. If you build something with it, star the repo and show us.
This is not a QVAC product, and it is not an accounting, bookkeeping, or tax product. It is an illustrative example, provided "as is", to show what a local AI app can do with the QVAC SDK. A local model will misread documents: the flags and the arithmetic check catch a lot, not everything, and every figure must be checked by a human before it is filed or paid. You alone are responsible for how you use it, including complying with the accounting and tax rules that apply to you.

A phone or an 8 GB laptop handles most everyday AI jobs, and video needs 32 GB. How much memory local AI needs, and the command that tells you what yours has.
Read More
We built Genesis III to push small models as far as they can go on STEM. A 1.7B model trained on it answers more than twice as many questions correctly as the same model trained on the open alternative, from an identical token budget.
Read MoreTranslatePsy-AfriSLM is our translation model for 19 African languages, small enough to run on a phone. We built a demo on it: paste text or photograph a page, and read it in your language, with no connection and nothing to pay.
Read MoreStay updated
New versions, breaking changes and migration notes - straight to your inbox. No spam, unsubscribe anytime.