ubereats

Uber Eats (ubereats.com)

TL;DR

curl -sL --compressed <category_url> returns a fully rendered HTML page with every store card in the DOM. Parse with cheerio. Pagination is ?page=N, up to 4 pages (~21 stores/page) in a typical city/cuisine combo. No anti-bot defeats required.

Detection

Signal	Value
CDN	Cloudflare (cf-ray, cf-cache-status: HIT)
Framework	Custom SSR — React hydration payload in <script type="application/json" id="__REDUX_STATE__">
Anti-bot	None on /category/ pages via curl. __cf_bm cookie is set but not enforced.
Auth	Not required for SEO category pages
robots.txt	Allows /category/

Architecture

Uber Eats ships a React SPA, but the /category// routes are fully server-rendered for SEO. Every store card is in the initial HTML as:

<a data-testid="store-card" href="/store/<slug>/<storeUuid>">
  <h3>Store Name</h3>
  ...
</a>

The parent div wraps: image (+srcset), name, ETA ("25 min"), price tier ("$", "$$"), optional promo badge ("Spend $35, Save $7"), and a "New" label (see Gotchas).

A window.__REDUX_STATE__ blob is also present but avoid it — it's embedded as a JSON-encoded string (with \u0022 for quotes) and contains a nested metaJson field that uses URL-encoding (%5C for backslash) which trips any stock JSON parser. Cheerio on the DOM is far cleaner.

Strategy used

• Phase 0 (curl) ✅ — HTTP 200, 1.2 MB HTML, all 21 store cards per page rendered inline. Done.
• Phase 1 — skipped (Phase 0 gate).
• Phase 2 — not needed.

Fetch — the one gotcha

Uber's content-security-policy header is ~40KB of whitelisted domains. Node's built-in fetch (undici) throws UND_ERR_HEADERS_OVERFLOW because its default header size cap is 16 KB. Two options:

1. Shell out to curl (simplest, what this skill uses).
2. Or use undici with a custom Agent({ maxHeaderSize: 65536 }).

import { execFile } from 'node:child_process';
import { promisify } from 'node:util';
const execFileP = promisify(execFile);

async function fetchPage(url) {
  const { stdout } = await execFileP(
    'curl',
    ['-sL', '--compressed', '-A', UA, '-H', 'accept-language: en-US,en;q=0.9', url],
    { maxBuffer: 32 * 1024 * 1024 }
  );
  return cheerio.load(stdout);
}