In the previous lesson we learned about the streamText function, capable of streaming in the response of an LLM.

In a terminal scenario, streaming text is basically all that we need because at the end of the day that is what we’re capable of showing. But we’re not here only to leverage the power of AI SDK in a terminal, we want to power our UIs!

Previously we leveraged textStream from the returning object of streamText, but this time we want to use toUIMessageStream(). A method that’s capable of returning so much more information about the LLM response.

import { google } from '@ai-sdk/google';
import { streamText } from 'ai'; // <- This is what you're looking for!

const model = google('gemini-2.5-flash-lite');
const prompt = 'Give me a sonnet about a cat called Steven.';

const stream = streamText({ model, prompt }); 

for await (const chunk of stream.toUIMessageStream()) {
  console.log(chunk);
}

If you paid close attention, beside the new toUIMessageStream() call, we have another change…

We went back using console.log instead of process.stdout.write, and that’s because while before we were getting a simple string and it was important to keep new words inline with the previous, the responso of toUIMessageStream() is quite different:

{ type: 'start' }
{ type: 'start-step' }
{ type: 'text-start', id: '0' }
{ type: 'text-delta', id: '0', delta: 'A' }
{ type: 'text-delta', id: '0', delta: ' feline friend,' }
{
  type: 'text-delta',
  id: '0',
  delta: ' with fur of purest white,\nHe stalks the halls, a prince of silent'
}
{ type: 'text-end', id: '0' }
{ type: 'finish-step' }
{ type: 'finish' }

Each object is a new chunk of our response, and the developers behind AI SDK decided this approach to give us plenty more information about “where we are at” in the current stream and thanks to the useful type values we can easily understand when the text-start, when it ends (text-end) and also how to build all the part in the middle since we are inside an iterable response and we can leverage the text-delta order to build the message in our UI.