How do I rate-limit OpenAI calls per user?

Why standard rate limiters are wrong for OpenAI

Tools like express-rate-limit or @upstash/ratelimit cap by IP over short windows. They cannot do "this user is on the free plan and has used 100 of 100 monthly renders." For AI apps, what you actually want is a plan-aware long-window quota tied to a user ID, with atomic enforcement under concurrency and refunds when the AI call fails.

AIPricingLab pattern

Reserve before the call, commit on success, release on error. Auto-releases after 60 seconds if your code crashes.

const r = await vevee.reserve(userId, "openai.chat", 1);
if (!r.allowed) {
  return new Response(JSON.stringify({ error: "rate_limited" }), { status: 429 });
}
try {
  const res = await openai.chat.completions.create(/*...*/);
  await vevee.commit(r.reservationId!);
  return Response.json(res);
} catch (e) {
  await vevee.release(r.reservationId!);
  throw e;
}

Stack with edge rate limits

For sub-second per-IP burst protection, keep a Cloudflare or Vercel rate limit at the edge. Use AIPricingLab for plan-aware long-window quotas. They coexist cleanly.