Models, prompts and routing configuration
The model, prompt and tool configuration each AI capability runs on, and how a change to it is released.
Note: This screen has no data behind it in this build
The address is specified and the console is ready for it. What is missing is the server side: the reads below are declared by the contract and are not registered in this deployment, so they answer “not implemented” rather than returning anything. Nothing on this page is a placeholder for data — there is no data to place.
What this screen is for
Each AI capability's configuration as a versioned record — model, prompt, tools, parameters — with the same draft, validate, publish and roll back sequence the catalog uses, so that a prompt change is a released version rather than an edit in place.
What has to exist first
- Reads
- staffAiConfigsList
- Actions
- staffAiConfigCreate
- staffAiConfigValidate
- staffAiConfigPublish
- staffAiConfigRollback
Each of those is an operation id in the frozen contract. Registering one is a server change, not a permission you can be granted — no group and no scope makes this screen return data today.