AI Gateway: Configuration Loader Policy
The new Configuration Loader policy splits loading an AI Gateway app’s configuration from running the app’s policy chain. Place it ahead of the Configuration Executor on your AI Gateway route, and any route policy you put between the two can read or adjust the loaded configuration before the app’s chain runs.
The loader is optional. Routes that list only the Configuration Executor keep working, because the executor still loads the configuration itself when nothing has loaded it yet. New AI Gateway projects include both policies on the scaffolded route.
Authentication placement stays flexible. Add API Key Authentication to an app’s own policy chain to require a key for that app only, or put it on the route before the loader to require a key for every app on the route.
Faster pickup of configuration changes
Both the Configuration Loader and the Configuration Executor take a
cacheTtlSeconds option that sets how long the gateway caches an app’s
configuration. The minimum is now 2 seconds, down from 10, so changes you make
to an app in the Zuplo Portal reach live traffic sooner. The default stays 10
seconds, and new AI Gateway projects ship with 5.
Getting started
See the Configuration Loader policy reference for setup steps, and Policy chains for how the loader and executor fit into an app’s chain.