# Caching & AOT Compilation Production deployments can precompute everything Kinetis would otherwise derive through reflection on every request — routing, MCP tool/resource registration, command registration, HTTP and MCP parameter binding, and DTO validation plans — into a build-time artifact, avoiding that cost at request time. This page covers what gets cached, how it's structured, and what it's worth in practice. ## `APP_ENV` ```{code-block} php use Kinetis\Runtime\AppEnvironment; $env = AppEnvironment::detect(); // reads getenv('APP_ENV') ``` ```{code-block} bash APP_ENV=development php -S localhost:8080 public/index.php ``` Unset or unrecognized values default to `Production`. Local development sets `APP_ENV=development` explicitly — see `.env.example` at the project root, or {doc}`config` for loading it from a `.env` file automatically. ## What gets cached Seven things get precomputed: - The route table. - MCP tool/resource definitions. - Command definitions. - The `#[AsGlobalMiddleware]`-discovered class list, already priority-sorted. - The `#[Listener]`-discovered event listener list, grouped by event class, already priority-sorted. - HTTP and MCP parameter-binding plans (how each request's data maps onto your controller/tool method's parameters). - DTO validation plans. Environment configuration (`.env`, see {doc}`config`) is not part of this cache — changing it takes effect immediately, with no rebuild needed. The on-disk result is five independent files: ```{code-block} text .kinetis-cache/ ├── http.php routes + global middleware + HTTP binding plans + │ validation plans for DTOs reachable from HTTP routes ├── mcp.php tool/resource definitions + MCP binding plans + │ validation plans for DTOs reachable from MCP tools ├── commands.php command definitions ├── events.php event listeners, grouped by event class └── openapi.php the generated OpenAPI document ``` A DTO used by both an HTTP route and an MCP tool appears in both files. ## Two ways to build it ### `bin/kinetis build` — pre-warm ahead of deploy ```{code-block} bash php bin/kinetis build # Compiled routes, MCP tools/resources, commands, event listeners, and # OpenAPI cache written to .kinetis-cache/ ``` Run this as part of your deploy step. Routes, MCP tools/resources, commands, global middleware, and event listeners are all found by namespace — see {doc}`cli` for how. ### Lazy, on first request If `APP_ENV=production` and no cache exists yet, the very first request compiles and writes it — safely, even under concurrent PHP-FPM workers racing to be "first" against an empty cache directory. Every request after that, on any worker, just loads what's already there. Once the cache exists, live discovery never runs again: your `Http`/`Mcp`/`Console`/ `Events` classes, and any `#[AsGlobalMiddleware]`-attributed class, aren't reflected again until the cache is rebuilt with `bin/kinetis build`. ```{note} Pre-warming avoids exactly one thing: the extra compile-and-write cost on whichever request happens to be first. It doesn't make caching *itself* any faster — a cold `prod-lazy` deployment is consistently slower than a pre-warmed one for exactly that one request, a one-time tax pre-warming removes entirely. ``` ## Performance characteristics For a persistent worker (FrankenPHP's primary deployment target — see {doc}`runtime-adapters`), `Router::register()` only ever runs once regardless of caching, since boot happens once for the whole worker's lifetime. What still runs on **every single dispatch**, cached or not, is `Dispatcher`/`Hydrator`'s parameter-binding and validation-plan derivation — precomputing that specifically is roughly 11% faster for path-parameter routes and up to 30% faster for query-parameter and validated-body-DTO routes. For a boot-and-die runtime (PHP-FPM), the picture is different. Reflecting an already-compiled class's metadata is cheap and roughly constant per method, regardless of that method's body size — a controller's file size isn't the lever. What matters is that live mode's `Router::register()` has to autoload *every* registered controller class on *every* request, because it can't know which route will match until the whole table is built. The cached path never touches those files at boot at all — `Router::fromArray()` holds class names as plain strings — so only the one controller actually dispatched to gets autoloaded, in either path. For an application with many controller classes, this makes the cached path several times faster overall: controller-class count and file size are the real lever at scale, not DTO or binding-plan count. Cold-start time (container/process startup) dominates the application-level difference entirely at that scale, but a `prod-lazy` deployment is still consistently the slowest cold configuration, for the same reason as above: it's still paying the compile-and-write cost a pre-warmed deployment already paid ahead of time. ## Cache invalidation The cache is not automatically invalidated on redeploy. Rerun `bin/kinetis build` as part of your deploy step whenever your registrations or DTO shapes change. ## See also - {doc}`cli` — how MCP tools/resources and commands are found by namespace, with no registration file for either. - {doc}`runtime-adapters` — why this matters enormously for `FpmAdapter` and barely at all for `FrankenPhpAdapter`'s own boot cost specifically. - {doc}`routing-validation` / {doc}`mcp` — the live behavior this precomputes; nothing about how routes/tools/DTOs are *declared* changes. - {doc}`persistence` — the *other* "cache" in this codebase: `Psr\SimpleCache\CacheInterface`, a general-purpose runtime cache unrelated to the build-time compilation described on this page beyond the shared word.