*`PIKU_AUTO_RESTART` (boolean, defaults to `true`): Piku will restart all workers every time the app is deployed. You can set it to `0`/`false` if you prefer to deploy first and then restart your workers separately.
*`NODE_VERSION`: installs a particular version of node for your app if `nodeenv` is found on the path. Optional; if not specified, the system-wide node package is used.
*`UWSGI_IDLE` (integer): set the `cheap`, `idle` and `die-on-idle` options to have workers spawned on demand and killed after _n_ seconds of inactivity.
> **NOTE:** `UWSGI_IDLE` applies to _all_ the workers, so if you have `UWSGI_PROCESSES` set to 4, they will all be killed simultaneously. Support for progressive scaling of workers via `cheaper` and similar uWSGI configurations will be added in the future.
> **NOTE:** if used with Cloudflare, `NGINX_HTTPS_ONLY` will cause an infinite redirect loop - keep it set to `false`, use `NGINX_CLOUDFLARE_ACL` instead and add a Cloudflare Page Rule to "Always Use HTTPS" for your server (use `domain.name/*` to match all URLs).
When `NGINX_CACHE_PREFIXES` is set, `nginx` will cache requests for those URL prefixes to the running application (`uwsgi`-like or `web` workers) and reply on its own for `NGINX_CACHE_TIME` to the outside. This is meant to be used for compute-intensive operations like resizing images or providing large chunks of data that change infrequently (like a sitemap).
The behavior of the cache can be controlled with the following variables:
*`NGINX_CACHE_PREFIXES` (string, comma separated list): set an array of `/url` values that will be cached by `nginx`
*`NGINX_CACHE_SIZE` (integer, defaults to 1): set the maximum size of the `nginx` cache, in GB
*`NGINX_CACHE_TIME` (integer, defaults to 3600): set the amount of time (in seconds) that valid backend replies (`200 304`) will be cached.
*`NGINX_CACHE_REDIRECTS` (integer, defaults to 3600): set the amount of time (in seconds) that backend redirects (`301 307`) will be cached.
*`NGINX_CACHE_ANY` (integer, defaults to 3600): set the amount of time (in seconds) that any other replies (other than errors) will be cached.
*`NGINX_CACHE_CONTROL` (integer, defaults to 3600): set the amount of time (in seconds) for cache control headers (`Cache-Control "public, max-age=3600"`)
*`NGINX_CACHE_EXPIRY` (integer, defaults to 86400): set the amount of time (in seconds) that cache entries will be kept on disk.
*`NGINX_CACHE_PATH` (string, detaults to `~piku/.piku/<appname>/cache`): location for the `nginx` cache data.
> **NOTE:** `NGINX_CACHE_PATH` will be _completely managed by `nginx` and cannot be removed by Piku when the application is destroyed_. This is because `nginx` sets the ownership for the cache to be exclusive to itself, and the `piku` user cannot remove that file tree. So you will either need to clean it up manually after destroying the app or store it in a temporary filesystem (or set the `piku` user to the same UID as `www-data`, which is not recommended).
Right now, there is no provision for cache revalidation (i.e., `nginx` asking your backend if the cache entries are still valid), since that requires active application logic that varies depending on the runtime--`nginx` will only ask your backend for new content when `NGINX_CACHE_TIME` elapses. If you require that kind of behavior, that is still possible via `NGINX_INCLUDE_FILE`.
Also, keep in mind that using `nginx` caching with a `static` website worker will _not_ work (and there's no point to it either).
