Warreniby Auton Systems

Ask your files. Get the page and line.

Warreni runs on your own computer, reads your folders — Word, PDF, Excel and images — and answers your questions from them. Every sentence carries a number that opens the original file where the answer came from. If the answer is not in your files, it says so.

Walkthrough on a call

Warreni

Your files, your language, your machine.

Nothing is uploaded. Warreni indexes the folders you point it at and keeps the index on your own machine.

  1. Point it at your folders

    Contracts, invoices, lectures, manuals. It reads each file once in the background, and again only when the file changes.

  2. Ask in Arabic or English

    Your question is searched in both languages, by the words and by the meaning, so an Arabic question finds an English file and the other way round.

  3. Open the source

    A model on your computer writes the answer from the passages it found. Each [1] opens the file at its page or line, so you check before you rely on it.

A real answer, not a mock-up.

Asked on a Windows PC, from a folder of machine-learning books. Exactly as it came out.

Warreni answering a question about vanishing gradients, with numbered sources and page numbers
Warreni on Windows with Qwen3-VL 8B, 27 September 2026: answered in 23 seconds from 84 PDF books.
  1. Your question, your languageThe books are in English; the question can be in Arabic. It is searched in both.
  2. An answer from the passagesWritten only from what was found, not from what the model remembers.
  3. Numbered sources, with pagesEach number is a book and a page. Click it and the PDF opens there.
  4. Written on the machineThe model and the time it took are shown under every answer.

It reads what your office already keeps.

PDF

The text of every page, with its page number, Arabic in the right order. Scanned pages are read while the computer is idle.

Word

Headings, paragraphs, tables and the pictures inside.

Excel and CSV

Every sheet as a table, under its own name, so a figure comes back with the sheet it lives on.

Images

A photographed invoice or a diagram is described by a vision model on your machine, and the description becomes searchable.

Private by where it runs.

Set up for the work you do.

A ready template for each kind of folder sets the examples, the filters and how the model answers.

Legal, finance, study, engineering

Templates for contracts, accounts, lectures and research, technical documentation, or a general folder — one per source.

Ask from your phone

A private Telegram bot answers with the same citations while the computer runs at home or in the office.

Follow up, save, act

A second question keeps the context of the first. Practical recommendations after each answer, and any result saved as a file.

Arabic, properly

Hamza and diacritics normalised, Arabic PDF text in the right order, a right-to-left interface — and the same app in English with one click.

What it needs.

Without a graphics card, a local answer takes tens of seconds rather than a few; text search is instant either way.

Questions people ask first.

Does it need the internet?

No. With its outside channels switched off it searches and answers fully offline. The internet is used only for the channels you leave on.

How is it different from a chatbot?

A chatbot answers from what it was trained on. Warreni answers only from your files, shows where each sentence came from, and says so when the answer is not there.

Can it read scanned documents?

Yes. Pages that are pictures with no text are marked, then read automatically while the computer is idle — or at once, with one button.

Can a whole office use it?

It runs on one computer and can be opened from others on the same network with an access code — and from outside through a private tunnel you set up.

How much does it cost?

Pricing is not published yet. Book a call and we will walk you through it on your own kind of files.

See it answer from your files.

Book a call. We run Warreni on a screen share and ask it the questions you ask every week.