Rate limits

Silktide counts the requests each key makes to each path. A key can make 75 requests to the same path in five seconds. The count starts with the first request and resets five seconds later, whatever happens in between.

The path is the whole path without the query string, so /v1/websites/2 and /v1/websites/3 are counted apart, and /v1/results/scores?websiteId=2 and /v1/results/scores?websiteId=3 are counted together.

When you hit it#

The 76th request gets 429, with a Retry-After header giving the seconds until the count resets:

HTTP/1.1 429 Too Many Requests
Retry-After: 3
Content-Type: application/json

{
  "error": {
    "type": "TooManyRequests",
    "message": "You have exceeded the maximum number of requests allowed in a short period of time. Please try again later.",
    "properties": { "retryAfter": 3 }
  }
}

Wait that long and carry on. The body is the same shape as other errors, and properties.retryAfter repeats the header. If the count itself cannot be checked, Silktide answers 503 of type ServiceUnavailable the same way, with a Retry-After of a few seconds.

async function silktide(url, init = {}) {
  for (;;) {
    const response = await fetch(url, {
      ...init,
      headers: { Authorization: `silktide token=${process.env.SILKTIDE_API_KEY}`, ...init.headers },
    });
    if (response.status !== 429 && response.status !== 503) return response;
    const seconds = Number(response.headers.get("Retry-After")) || 5;
    await new Promise((resolve) => setTimeout(resolve, seconds * 1000));
  }
}

Staying under it#

• Ask for what you need once. List websites and their sections answers for every website in one request.
• Poll slowly. A website's test takes minutes at least, so checking it every few minutes is plenty.
• Use one key per service. Keys that share a job do not share a limit, but spreading one job over many keys to dodge it can get them blocked.