Skip to main content
Every request is live: Clair loads the page when you ask, so the response time is mostly the time the source site takes to hand over its data. Some sites answer a plain request in a second or two. Others only show their data inside a real browser, after a wait, a click, or a scroll. Those take longer on every request, and that time is expected. Each endpoint is labeled by how it loads, with the time most requests finish in and the limit after which Clair stops and returns an error.

How an endpoint loads

Reading the numbers

  • Typical is the range most requests finish in, measured on live requests with nothing cached. A request inside it is working as intended.
  • A request above the range but under the limit usually means the source site blocked the first try and Clair tried again a different way. This happens on a small share of requests to any site.
  • Limit is where Clair stops. The request returns 504 and is not charged. Retry it.
  • A repeated request can come back in milliseconds: Clair keeps a result for a short time and serves the same request from it.
Set your client timeout a few seconds above the endpoint’s limit, so you get Clair’s error body and not a dropped connection. To get through a long list faster, send requests in parallel within your plan’s rate limit; one slow request does not hold up the others. A 503 overloaded means the endpoint is at capacity for the moment: wait a second and retry.

Every endpoint

Web Scraper API

Amazon API

Google Search API

Google Maps API

ChatGPT API

Google AI Overview API

Google AI Mode API

YouTube API

TikTok API

Instagram API

Glassdoor API

Indeed API

Google News API

Google Shopping API

Trustpilot API

Crunchbase API

Reuters API

Bloomberg API

Associated Press API