Minimum
Demo and edge
- What it needs
- 16–32 GB RAM, no GPU required. 4+ CPU cores and 60 GB of disk for the installer itself.
- Fits
- Trying ReduceWaste, small installs and the translation pipeline.
Deploy and size sites for your clients.
From a single-box demo to a multi-line plant split across several boxes by role: historian and app, local inference and camera vision. The same engineer-facing surface works across every site you administer.
Two hardware tiers
Recommended
Baseline
- What it needs
- Either a GB10-class unified-memory box (128 GB shared, 96 GB or more of it free) or discrete GPUs with at least 48 GB of combined VRAM.
- Fits
- The full multi-step analysis workload. The reference box, with 2× RTX 3090 and 48 GB combined, is sized at this tier.
Sizing by site size, and what changes for multi-line plants, is in the hardware guide.
What you take to a client
A plan from their files
Photos, drawings and manuals go in. A monitoring plan and a vendor-neutral sensor shopping list come out, with the reason for each sensor.
Read-only on the plant
Adapters are read-only toward the plant. Modbus write function codes are rejected in code and OPC UA sessions are opened read-only. The edge reader is the only component that talks to field devices.
Every site, one surface
The engineer-facing screens are the same at each client, so what you learn at one site carries to the next.
Install with Docker Compose Air-gapped install OT safety and IEC 62443
Tell us about your sites.
Say how many client sites you look after and what you want to deploy. We reply by email.
Enquiry received
Thank you. We have your details and will reply by email.