Skip to content

Rate limits

The Gib engine has no request limit. Nothing in it counts or throttles requests as they arrive. An operator can still put a proxy or a content delivery network in front of an instance, and that layer can set its own limits.

The image-resize pipeline has a rate limiter. It guards against a flood of resize variants filling the cache. It does not limit how many times you can fetch an image.

These limits are built into the engine. Each server process allows:

  • 5 new variants per image per minute.
  • 100 new variants per minute across all images.

When a budget is spent, Gib still sends you the resized image. It only does not store the new variant. The limits count per server process.

A daily job removes variants that were used fewer than three times in the last 24 hours.

This applies most to the public instance at gib.show, which the project pays to run. Be reasonable. A heavy user should cache with the long-lived, immutable /image/direct/{imageHash} route. Do not fetch the same token or chain image again on every request. See Caching.

Request limits may be introduced later on the public instance. This page will say so plainly if they are.