Rate Limits

Rate limits protect the platform and ensure fair capacity allocation across all accounts. Limits are enforced per API key and reset on a rolling per-minute window.

Per-key rate limits#

The default limit is 60 requests per minute per key. If you need a higher limit, you can configure it per key from your dashboard once your account is verified.

Keys created programmatically via POST /api/keys can also have a rate_limit_rpm set at creation time.

Chat limits#

Chat conversations on opendunes.com/chat are limited to 30 messages per minute per account, on every model including free ones. The limit exists to keep free models responsive for everyone; normal conversation never approaches it.

Note. Limits are enforced at the gateway level, before a request reaches any model provider. A request that exceeds your limit returns immediately without consuming any credits.

Rate-limit headers#

Every response includes three headers that tell you your current limit state:

X-RateLimit-Limit-RPM: 60
X-RateLimit-Remaining-RPM: 43
X-RateLimit-Reset: 1750000060

Read X-RateLimit-Remaining-RPM before sending bursts to avoid hitting the wall mid-batch.

Handling 429#

When you exceed the limit, the server returns 429 Too Many Requests with a Retry-After header and a matching retry_after field in the JSON body:

{
  "error": "rate_limit_exceeded",
  "message": "Rate limit: 60 requests per minute.",
  "retry_after": 8
}

Recommended strategy:

1. On 429, read retry_after (or Retry-After) from the response.
2. Wait that many seconds plus a small random jitter (e.g. retry_after + random(0, 2)).
3. Retry the original request unchanged.

For high-throughput pipelines, pre-check X-RateLimit-Remaining-RPM from the prior response and proactively slow down before hitting 429.

Credits remaining#

The X-Balance-Available header is present on every successful response and reports your remaining balance in micro-DA (1 DA = 1,000,000 micro-DA):

X-Balance-Available: 4832000000

4,832,000,000 micro-DA = 4,832 DA remaining. Use this to surface low-balance warnings in your application without polling a separate balance endpoint.

When the balance reaches zero, subsequent requests return 402 insufficient_credits. Top up via your dashboard using a CIB or Edahabia card.