How can users send a new message after a streaming error?

[jin-mint]

Hello, I’m building an internal service with ai-chatbot,
and first of all—thank you for releasing such a great open-source project. 🙌
I opened this issue because I’m unsure how to let users enter a new message after an error occurs during streaming.

Description
In the multimodal-input component the user can only submit a new message when status is 'ready'.

While a response is streaming, if an error occurs the status is set to 'error'.
At that point pressing Enter no longer triggers a submission, so the user is stuck until the page is refreshed.

I would appreciate guidance on the best-practice way to let the user send another message after an error.

// multimodal-input.tsx

      <Textarea
        data-testid="multimodal-input"
        ref={textareaRef}
        placeholder="Send a message..."
        value={input}
        onChange={handleInput}
        className={cx(
          'min-h-[24px] max-h-[calc(75dvh)] overflow-hidden resize-none rounded-2xl !text-base bg-muted pb-10 dark:border-zinc-700',
          className,
        )}
        rows={2}
        autoFocus
        onKeyDown={(event) => {
          if (
            event.key === 'Enter' &&
            !event.shiftKey &&
            !event.nativeEvent.isComposing
          ) {
            event.preventDefault();

            if (status !== 'ready') { // this point
              toast.error('Please wait for the model to finish its response!');
            } else {
              submitForm();
            }
          }
        }}
      />

Approaches I’m considering

1. Automatically reset status
   – After setting status to 'error', dispatch another state update that flips it back to 'ready' after a short delay.

2. Loosen the keyboard guard
   – Change the onKeyDown condition to allow submission when status === 'error':

if (status === 'ready' || status === 'error') {
  submitForm();
}

[0Chan-smc]

Still looking for a solution here—I’m experiencing the same issue.

[DaveInga]

i'm implementing a custom chat API endpoint in Lambda and experiencing the same error. Can't test a second request until the status is ready.