Lots of changes in this new version, with some major rework in the code itself. Hence version 2.0.0!

1. Improved influx writes for metric logging
2. robots.txt is now properly handled on remote CDN as well
3. Database metrics are exposed
4. Configurable database pool size
5. Migrations of database entries are now handled for you
6. Oversized image upload requests are now properly handled
7. Proper handling in Docker with dumb-init

It has been a while since we last released a new version from the beta channel, but a new one is coming. Most notably, this version includes support for WEBP images. This means that WEBP enabled clients can benefit from images with a better quality/compression ratio!

How to enable webp?

WEBP is a newer format introduced by Google. It's goal is to further reduce the size of an image, but maintain quality. By now, quite a number of people have a browser which support webp, so it may make sense to enable it.

Test it

The example default.json.example contains a new webp section. When the allow_opt_in is set to true, clients can opt-in for webp. iaas will determine whether a client understands webp. If they do, and they have opted in, a webp image will be sent back. Clients opt in by setting the following query parameter in the request allow_webp=true.

You can modify your application to send this param for just a few users to test out behaviour, and let iaas populate its caches with webp content.

General roll-out

Once you are confident webp works for your case, you can enable it by setting the following flag in the config to true: allow_dynamic_switch (under webp). That will allow all clients which accept webp to pick this over png or jpeg. Opting in is not required.

Note on quality

The quality options between webp and jpeg are not identical. Try a proper quality setting for your environment before committing to serving webp to all of your clients.

Other fixes

• Now using node 8.9
• Clearer error upon token race
• Strip comments and other non useful data from the image
• Improved buffer handling

• Make the database layer pluggable
• Hard disconnect from the database if it does not come online

This release the focus was on two mean features: proper metrics support and quality support for JPG

So the changelog goes as follows:

• Support a quality=X (where X is an integer between 0 and 100) to indicate the intended quality of an image
• Support to send lots of metrics to a database as influx (using an extensible layer)
• Improved timeout handling of conversion processes

Some new features warranted a new release! We have tried to make stuff even better for you :)

• Images are now properly flattened
• Conversion processes timeout after some time to prevent a meltdown of the service
• On the loss of a database connection, the system will restart and reconnect

We are proud to present the first tested stable version of 'Imaging-As-A-Service'.

This version has been running in production at both Magnet.me and inventid as has proven itself as a fast, scalable, and reliable service. It quickly generates different types of images based on the requested URL. As a result, other services do no longer need to handle any type of image manipulation at all!

Changes since v0.1.11:

• Addition of cache control on redirect headers (Thanks @TiddoLangerak)
• Add a cover mode to the requested image scale types
• Improved stream handling for disconnecting clients while a conversion is in progress
• Limit the maximum image size (in pixels) for uploaded images to 30 Megapixel

Try using docker:

docker pull inventid/iaas:release-v1.0.0

This release fixes a bug where interlaced JPEGs were uninterlaced.

This new release adds the rendered_at parameter to the AWS S3 cache name. This prevents race conditions in case one needs to clear some caches.

This release adds improved monitoring of the performance in the the logs

Bugfix release:

• This make the cropping on upload more reliable when presented with falsy values