What is this

It's an attempt at building a UI framework powered by an LLM, that is The LLM decides what to display, based on the following:

• Current State
• User Input
• Other Context

How does it work

Right now

• We pass the UI Schema (ie: Available widgets) as context to the LLM
• We pass the current state as JSON
• The LLM is instructed to create a new state based on the input (Including the user instruction)

Problems with the approach

• Scalability, the more "widgets" we have the more context we need

Next approach

TBD

How can I run it

• make be # This will run the backend (Websocket server)
• make fe # This will run the frontend (Static site, that connects to websocket)