# Caching

Two things are cached through your app's default cache store via the `Cache` facade. The package does not configure or require a dedicated cache store.

## What Is Cached

- Model schema, including columns, casts, and enum values, is built once per model and cached for one hour.
- Prepared requests, meaning the parsed and validated query plan keyed by a fingerprint of the request plus the model config, are cached for five minutes.

The prepared-request cache exists mainly for approvals. A request that needs approval is evaluated twice, once to decide whether to pause and once to actually run after approval, and this cache prevents the second pass from re-parsing and re-validating from scratch.

## Cache Keys And Invalidation

Both caches use versioned, `eloquent-ai-tools:`-prefixed keys, so sharing a cache store with the rest of your app, or other apps, is safe.

The model-schema cache is refreshed on use by touching its TTL when a cached snapshot is read.

Both caches are invalidated automatically after `php artisan migrate` via a listener on `MigrationsEnded`.

If you change a model's schema some other way, such as a raw `ALTER TABLE` or by editing `$visible`, `$hidden`, or casts on the model, and do not want to wait for the TTL, call `ModelSchemaBuilder::clearCache()` yourself. That also clears the prepared-request cache.

## Production Note

If your app's default cache store is the `array` driver, neither cache persists across requests. Everything still works correctly, but you do not get the intended performance benefit.

Use a persistent store such as `file`, `database`, or `redis` in production to benefit from caching.