Prerequisites
Proposal
AI dynamically generating or adapting interfaces based on user intent
There is two parts to this. The generation: ability to process the user prompt generate UI based on the intent and the building blocks to use for generating interfaces. I can see how sgds currently can cover the building blocks part, but the generation based on user intent is another layer on top of building blocks which we have not explored.
AI helping users complete workflows (like form filling or applications)
This part, I think I dont really understand this. I am not sure how is this linked to the design system.
Let me try to process this GenUI....
Case 1: User of a product
i.e. "I want to open a F&B stall" --> triggers AI to pull out checklist of todos --> its a checklist so AI use SgdsCheckbox component to generate the list
I do see similarities in our current workflow, except that the prompts are different because our users are the developers.
Case 2: Developing a product (currently supported by SGDS)
Developer users (designers prototyping or engineers building a site) :
i.e. "I want to build a multi step form" --> triggers AI to use SgdsStepper and Sgds form components
The difference is the prompts are of differing nature. Case 1 will be more product specific case to configure prompt trigger that can take i want to open a F&B stall --> generating checklist UI
Another product might take a different generation for their use case I want to apply for grant --> generated form application UI . I want to register marriage --> Step by step guide in the form of Cards
**I can see how SGDS can come in to provide building blocks for the UI and the guides. Breaking down the example futher:
i want to open F&B stall --> Product finds what to do with this and return a list --> SGDS says list should use static SgdsDescriptionList , if its a sequential list then SGDS says use SgdsNumberedList
The building blocks yes we have. The relationship logic for composition of UI no we don't have that yet. When I was writing the agent layer for the DS, we have to write rules of usage.. which i feel is a similar concept to this relationship logic.**
Motivation and context
Prerequisites
Proposal
AI dynamically generating or adapting interfaces based on user intent
There is two parts to this. The generation: ability to process the user prompt generate UI based on the intent and the building blocks to use for generating interfaces. I can see how sgds currently can cover the building blocks part, but the generation based on user intent is another layer on top of building blocks which we have not explored.
AI helping users complete workflows (like form filling or applications)
This part, I think I dont really understand this. I am not sure how is this linked to the design system.
Let me try to process this GenUI....
Case 1: User of a product
i.e. "I want to open a F&B stall" --> triggers AI to pull out checklist of todos --> its a checklist so AI use SgdsCheckbox component to generate the list
I do see similarities in our current workflow, except that the prompts are different because our users are the developers.
Case 2: Developing a product (currently supported by SGDS)
Developer users (designers prototyping or engineers building a site) :
i.e. "I want to build a multi step form" --> triggers AI to use SgdsStepper and Sgds form components
The difference is the prompts are of differing nature. Case 1 will be more product specific case to configure prompt trigger that can take i want to open a F&B stall --> generating checklist UI
Another product might take a different generation for their use case I want to apply for grant --> generated form application UI . I want to register marriage --> Step by step guide in the form of Cards
**I can see how SGDS can come in to provide building blocks for the UI and the guides. Breaking down the example futher:
i want to open F&B stall --> Product finds what to do with this and return a list --> SGDS says list should use static SgdsDescriptionList , if its a sequential list then SGDS says use SgdsNumberedList
The building blocks yes we have. The relationship logic for composition of UI no we don't have that yet. When I was writing the agent layer for the DS, we have to write rules of usage.. which i feel is a similar concept to this relationship logic.**
Motivation and context