Sovorix runs inside your plant, only reads your data, and never acts without a person's approval. Here's how, in plain words first and in detail for your security team below.
Sovorix runs on servers at your site, including the AI models. It works with no internet connection. Your plant data, your questions and every receipt stay on your network.
In a first deployment, Sovorix connects to your historians, SCADA and cameras with read-only access. It has no path to write to your controllers (PLCs or DCS). It suggests; your people act.
Every suggestion goes to a named person whose role allows them to approve it. They can approve, hold or reject. The operator makes the change and confirms it, and both steps are recorded.
Each receipt records what was suggested, the evidence, who approved it and when. Receipts show any tampering: if anyone edits one later, the check fails. Try it below.
The check runs in your browser. Nothing is sent anywhere.
The detail, answered plainly. If something isn't covered, email udayan.pawar@artpark.in and we'll answer in writing.
Sovorix runs on your hardware at your site, on a server you control. For a one-line pilot, we size it at one server with a GPU (24 GB or more of GPU memory), 64 GB of RAM and 2 TB of storage. It can run on bare metal or in a virtual machine.
Sovorix opens read-only connections from its own server to the systems you allow: your historian, SCADA, camera recorder and maintenance system. It needs no inbound connection from the internet and no internet connection at all. People reach its web interface from inside your network. We give your team the exact list of ports and protocols during setup.
Sovorix sits at site operations level, alongside your historian (Purdue Level 3), or in your industrial DMZ (Level 3.5) if you prefer. It doesn't connect to control networks at Levels 0–2.
Connectors use read-only accounts and read-only interfaces, such as historian read APIs, OPC UA subscriptions and MQTT subscriptions. In a first deployment there's no write path to controllers. Changes on the line are made by people, through your existing systems.
All models run on the Sovorix server at your site. Nothing is sent to an outside AI service, and no plant data is used to train models for anyone else.
Each person signs in and sees only what their role allows. Roles decide who can view which lines, who can approve which kinds of change, and who can manage settings.
When someone approves a suggestion, Sovorix writes a receipt: what was suggested, the evidence, who approved and when. Each receipt is sealed with a cryptographic fingerprint and linked to the receipt before it, so any later edit, deletion or reordering shows up when the receipts are checked. Your team can check them at any time, with no help from us.
You decide how long Sovorix keeps readings, camera frames and receipts. By default, it keeps readings for 2 years, camera frames for 90 days, and receipts for as long as you need. When a pilot ends, we remove Sovorix from your server, and your data stays with you.
Our team works on site, alongside yours. There's no remote access to your network, and nobody from Sovorix works on your systems without your team's approval. The final report is written with you and holds only the agreed number, how it was measured and what we found. You review it before any copy leaves your site.
Sovorix only suggests. Every suggestion shows the evidence it's based on, so your team can see when it doesn't add up and reject it. Nothing changes on the line unless a person approves it and an operator makes the change.
Contact for security questions: udayan.pawar@artpark.in
