AI SDK - Am I going insane? - v0

Literally cannot get it to work all day, no matter whether I use the vercel agent otr the gemini agent, every single respnse comes back with an error

Even a simple app with a text box and a generate button, and the EXACT example from the documnetation produces an error.

1import { generateText } from "ai"23import { openai } from "@ai-sdk/openai"45678export async function POST(req: Request) {910  try {1112    const { prompt } = await req.json()1314151617    if (!prompt) {1819      return Response.json({ error: "Prompt is required" }, { status: 400 })2021    }2223242526    const { text } = await generateText({2728      model: openai("gpt-5"),2930      prompt: prompt,3132    })3334353637    return Response.json({ text })3839  } catch (error) {4041    console.error("Error generating text:", error)4243    return Response.json({ error: "Internal Server Error" }, { status: 500 })4445  }4647}

I have the same OpenAI api key in place in other places, and so that’s not the issue - What on earth is going on with this?

https://v0.app/chat/generate-email-with-supabase-bYd23AOAZpJ

Your try/catch is wrapping the entire function so anything that errors inside it is going to trigger the same error. In this case it's not the AI Generation that is failing

"Invalid JSON Response" is thrown when you try to use either Response.json() or request.json() on a body that is not JSON. Internally it's failing on the JSON.parse/stringify steps.

Try const body = await req.text() at the top of the handler to see what actual input body you're getting

, just checking in to see if you're still experiencing issues with the AI SDK. Have you had a chance to try the suggestion I provided to log the request body? Let me know if that helps!