GuidesAI development
Tutorial: streaming an AI response in a web interface
1. Server side
An endpoint that returns a text stream (example in Astro route format, easy to port to Express or Next.js):
import Anthropic from '@anthropic-ai/sdk';
const client = new Anthropic();
export async function POST({ request }: { request: Request }) {
const { question } = await request.json();
const stream = client.messages.stream({
model: process.env.CLAUDE_MODEL!,
max_tokens: 1024,
messages: [{ role: 'user', content: question }],
});
const enc = new TextEncoder();
const body = new ReadableStream({
start(ctrl) {
stream.on('text', (t) => ctrl.enqueue(enc.encode(t)));
stream.on('end', () => ctrl.close());
stream.on('error', (e) => ctrl.error(e));
},
});
return new Response(body, { headers: { 'Content-Type': 'text/plain; charset=utf-8' } });
}
2. Browser side
const res = await fetch('/api/ask', { method: 'POST', body: JSON.stringify({ question }) });
const reader = res.body!.getReader();
const dec = new TextDecoder();
for (;;) {
const { done, value } = await reader.read();
if (done) break;
output.textContent += dec.decode(value, { stream: true });
}
3. The details that matter
- An indicatorA blinking cursor while the stream is open.
- The Stop buttonAn AbortController, wired to the server-side stream.
- MarkdownRendered at the end, or block by block, to avoid jumps.
Before this: calling the Claude API from Node.js.
Frequently asked questions
- Does streaming cost more?
- No. The number of tokens is the same; only the way you receive them changes.
- Do I need WebSockets?
- No. An HTTP response read in chunks is enough, or Server-Sent Events.
- How do I stop an answer in progress?
- In the browser, an AbortController cancels the request; on the server, you stop the SDK stream when the connection closes.