render function error handling

[steebchen]

Description

It looks like the render function doesn't support error handling. I tried wrapping the render function in a try/catch, but it doesn't catch any errors (specifically when an error is immediately returned, such as a context window length exceeded)/

Code example

I used this template: https://github.com/vercel/ai-chatbot

At this part:

return render({
  model: 'gpt-3.5-turbo',
  provider: openai,
  initial: <SpinnerMessage />,
  messages: [
    {
      role: 'system',
      content: `You are a helpful bot.`
    },
    ...aiState.get().messages.map((message: any) => ({
      role: message.role,
      content: message.content,
      name: message.name
    }))
  ]
})

This can result in this error being thrown

 ⨯ unhandledRejection: BadRequestError: 400 This model's maximum context length is 16385 tokens. However, your messages resulted in 21253 tokens. Please reduce the length of the messages.

and the UI just keeps showing this:

Additional context

No response

[parthematics]

Also experiencing this issue - would love some guidance here!

[nateaxcell]

Same, its obvious that the openAI library is throwing an error but there's no consistent way to catch it using this library it seems.

[ghaemisr]

Same here! Couldn't find any documentation on how this should be handled.