ampproject/toolbox-optimizer repository overview

⁠ampproject/toolbox-optimizer

ampproject/toolbox-optimizer⁠ is an image that exposes an AMP Optimizer⁠ server for optimizing your amphtml using the same server-side-rendering optimizations as the Google AMP Cache⁠.

⁠How it works

The AMP Optimizer server accepts POST requests at port 3000 and requires an HTML body. It then runs the HTML through the optimizer⁠ package, and returns the result as the response.

⁠Configuration

There are two different kinds of configuration you can supply to the container:

1. Static configuration via environment variables: On startup, the container will search for environment variables prefixed with AMP_OPTIMIZER_ and pass the values as configuration options when initializing the underlying optimizer library. The full list of options are available here⁠. Options should be specified in SCREAMING_SNAKE_CASE as opposed to camelCase. For example, in order to configure the preloadHeroImage option, you would declare the environment variable named: AMP_OPTIMIZER_PRELOAD_HERO_IMAGE.
2. Per-request configuration: there are some options that can only be set on a per-request basis. For example, the canonical flag for specifying an AMP page's canonical link. These can be specified via query params, e.g. /?canonical=http://example.com.

⁠Usage

⁠Running a basic optimizer server

$ docker pull ampproject/toolbox-optimizer
$ docker run -it ampproject/toolbox-optimizer

⁠More complex configurations

The amp-toolbox-docker-optimizer image can be layered and composed using any of the usual container orchestration tools, like Docker Compose⁠ or Kubernetes⁠. An example of using Docker Compose is provided under the demo directory.

⁠Best Practice: Cache server-side-rendered AMPs

To achieve best performance, transformations shouldn't be applied for every request. Instead, transformations should only be applied the first time a page is requested, and the results then cached. Caching can happen on the CDN level, on the site's internal infrastructure (e.g.: Memcached), or even on the server itself, if the set of pages is small enough to fit in memory.