Generating with AI
Describe what you want; get a model back that has already been validated.
How it works
Open the AI panel from Generate with AI in the toolbar and describe the system:
A temperature and humidity sensor for a greenhouse, with a serial number and a calibration method.
What comes back is not a first draft thrown over the wall. The generator:
- Detects the companion specifications your description implies — a robot arm pulls in Robotics, a machine tool pulls in MachineTool — so the model reuses standard types instead of inventing parallel ones.
- Generates the YAML.
- Compiles and validates it, and corrects its own errors, looping until the model is clean or it runs out of attempts.
The panel reports what happened — how many lines were generated, and how many errors remain if any survived the correction pass.
Refining
After the first result the input switches to Refine the model…. Follow-ups are applied to the model you already have, so you can work in steps:
Add an alarm when the temperature goes above 40 °C.
Move the serial number into the identification block.
This is usually better than one long prompt: each step is validated on its own, so when something goes wrong you know which instruction caused it.
What it is good at, and what it is not
Good at the shape of a model: which companion spec applies, which standard type to subtype, where a value belongs, what the naming conventions are. That is knowledge the generator has and most people are still acquiring.
Not a substitute for review. It produces a model that compiles; whether it matches the machine in front of you is yours to judge. Read the diagram — it is the fastest way to see whether the structure is right.
Requirements
Signing in is required. See Overview.
The same capability is available to AI agents through the MCP server's
opcua_model_create tool — see the
MCP connector guide — and over the
REST API.