Work is complete on the main toolset for the rendering machine. We’re using the Serverless Application Model to build and test everything offline, then deploy to AWS. There are two APIs – one internal to run the rendering workers, the other public facing to integrate with customer and control UIs in the Sanity headless CMS.

One key design requirement is that it should be possible to add a customer, and add new render formats for a customer, without editing the YAML template, and without redeploying the API. The only change should be to the configuration in the Sanity UIs and, if required, new code modules dropped in to Lambda functions. This is exactly how it works, and so we can build out for new customers without interrupting live work.

Now we’re ready to set up a UI for a guinea pig customer, and code the first stage of the engine. That’s the bit which collects data from any API anywhere, and deposits it into Sanity for moderation and onward delivery to second stage render workers running in the customer’s AWS account.