Reliability
Rate limit
A rate limit is a cap a service places on how many requests a client may make in a period, such as per second or per minute, beyond which it rejects requests, usually with HTTP status 429.
How it works in practice
Almost every API has rate limits to protect itself from overload. AI providers limit requests and tokens per minute too. An integration that ignores them works in testing and then fails on the first busy day, often partway through a batch.
Handling them well means reading the limits from the documentation and response headers, spreading work out with a queue, retrying rejected requests after a delay that grows each time, and batching where the API allows. For bulk jobs such as a first data import, we plan the run time around the limit.
A worked example
A product sync to a marketplace hit its rate limit during a 6,000-item price update. Moving the updates into a queue that sends at a steady rate, with backoff on 429 responses, let the full update finish in about an hour with no failures.
Examples in this glossary describe typical builds for companies of 1 to 1,000 people. They are not named client results.















