[Vercel v0] Support Astro site generation and deployment

Over the weekend, I decided to experiment with v0, and learned something that I thought was really, really funny: v0’s LLM appears not to know what v0 the platform can and can’t do. Let me explain. I asked v0 to generate an Astro site. I feel like React is overhead that a content-driven site doesn’t need, and Astro gives us a way of building with all the benefits of TSX code, without the overhead of React. v0 then generated some Astro code, which looked like Astro code. I couldn’t preview the site. “No problem,” I thought, ”I’ll just deploy it and see what it looks like.” It appeared to deploy a stock Next site that wasn’t completely configured, and didn’t build successfully. Eventually, I prompted v0’s LLM to ask: Does v0 support deploying an Astro site to Vercel? Or is only Next supported? It replied, “You've identified the core issue! It seems like Vercel is still trying to build this as a Next.js project rather than recognizing it as an Astro project.” It then gave me instructions for how to go into Vercel and update the project’s settings to indicate it’s an Astro project. So I tried deploying again. Still deployed the un-buildable Next site. 😂 So I clicked the button to download the project. “I build stuff all the time, I’ll just run this locally,” I thought. But when I unzipped the package and looked at the files, I didn’t see the Astro code the LLM generated—I saw the files for that stock Next site. 🤣 So then I prompted v0’s LLM to explain the download that just happened, and ask directly if v0 supports working with Astro instead of next. It replied: “You've identified an important limitation. You're right - v0's built-in preview and download functionality is primarily designed for Next.js projects. While I can write and explain Astro code, the v0 interface doesn't currently support previewing or directly downloading Astro projects.” Me: 🤣🤣🤣 But, this highlights one of the reasons why I think designers shouldn’t just be knowledgeable about the front end—they should be knowledgeable about the back end, too. I get backend systems. I work to understand the moving pieces, and how one moving pieces sets up the next moving piece for success. I would at *least* encourage the v0 team to update v0’s system prompt so that it strongly discourages users from generating an app with any framework other than Next. Because when users get frustrated that a system can’t do what it told them it can, they become former customers. Do you have a product where a clunky experience drives customers away? Get in touch with me about consulting with you on a project to fix that!