Every AI feature in Pyron — the assistant, the builders that draft forms, pages, reports and automations, and the AI steps inside automations — runs on a model in one of three tiers. You choose the model behind each tier and the tier each kind of work runs on.
Choosing models
The model settings live on the AI page, one of the areas of Setup. Open its Models tab.
The top of the view sets the model behind each tier:
- Standard model — most chats and building.
- Best model — the hardest building, when it is allowed.
- Fast model — titles, web reads and quick steps.
A tier offers a choice only when more than one model can serve it.
Under By kind of work, each row names one kind of AI work — such as Chat assistant, Form builder or AI steps in automations — with the tier and model it runs on. Pick a different tier on a row to move that work to the model behind that tier.
Your changes wait for Save in the header, and Discard returns to the saved settings. Saved changes apply from the next AI request.
To check that the AI answers, select Test connection. Pyron sends one short request to the model the chat assistant runs on and shows whether it answered or, if it could not, the reason it gave.
Where AI runs
Every AI request runs on Pyron's AWS Bedrock. There is no provider to choose and no key or account to connect: choosing a model is the whole setup.
Each model in a tier's list says where it runs. A model in Australia keeps your data in Sydney and Melbourne and costs about 10% more. A model that runs worldwide costs list price, and your data may leave Australia, so choose one only if that suits your organisation.
What the AI may handle is set in AI guardrails, and what it may spend in AI usage and budgets.