Is your feature request related to a problem? Please describe.
On your blog for future problems you mention context as the first big problem to solve. Several salient sources of information are brought up there except for one that I'm surprised more startups haven't tried to tackle yet. Frontend designs!
If I'm writing a component in react, obviously the biggest source of context I'm considering is the figma I'm looking at. It would be great if there was a way to either upload images or connect to popular tools like Figma directly and even @ them when writing code for them.
I realize there are some people in the space doing code generation off of figma (like builder.io and replit), but what those are missing is the context of our own codebase. What's in my tailwind file? How am I using something like headless UI? And so on....
I'm thinking mostly a copilot here, but long term I would even love if it could run and inspect components in isolation and determine UI bugs. When we get full file agentic generation, some kind of preview feature would be nice. Even including some interaction! But to start I just want to propose a simple copillot that would for example accurately predict tailwind styles based on the image (or get close)
Describe the solution you'd like
I upload a design image or connect to figma directly. Copilot++ nicely infers styles and other info based on the image and other important pieces of context, such as my tailwind.config file, ui libraries i'm using (such as headless UI), and other UI components I've done so it knows my style
Is your feature request related to a problem? Please describe. On your blog for future problems you mention context as the first big problem to solve. Several salient sources of information are brought up there except for one that I'm surprised more startups haven't tried to tackle yet. Frontend designs!
If I'm writing a component in react, obviously the biggest source of context I'm considering is the figma I'm looking at. It would be great if there was a way to either upload images or connect to popular tools like Figma directly and even @ them when writing code for them.
I realize there are some people in the space doing code generation off of figma (like builder.io and replit), but what those are missing is the context of our own codebase. What's in my tailwind file? How am I using something like headless UI? And so on....
I'm thinking mostly a copilot here, but long term I would even love if it could run and inspect components in isolation and determine UI bugs. When we get full file agentic generation, some kind of preview feature would be nice. Even including some interaction! But to start I just want to propose a simple copillot that would for example accurately predict tailwind styles based on the image (or get close)
Describe the solution you'd like I upload a design image or connect to figma directly. Copilot++ nicely infers styles and other info based on the image and other important pieces of context, such as my tailwind.config file, ui libraries i'm using (such as headless UI), and other UI components I've done so it knows my style