Fork me on GitHub

Introducing diecutter

Today, we initiated a proof-of-concept template generation service called diecutter:

  • at a given URL, there is a "template" resource;
  • post variables to resource's URL and receive the rendered string.

The resource is, basically, an engine (Jinja) and a template string.


The project is at a really early stage, but you can try it already.

Install diecutter from Github and run the server:

git clone
cd diecutter/
make develop
make serve &

Check it works:

$ curl http://localhost:8106
{"diecutter": "Hello", "version": "0.1dev"}

Put your template in the service templates directory or use the API:

$ echo "Hello {{ who }}" | curl -X PUT http://localhost:8106/hello -F "file=@-"
{"diecutter": "Ok"}

Then we can get the raw template we just created:

$ curl http://localhost:8106/hello
Hello {{ who }}

And we can render the template against some variables:

$ curl -X POST http://localhost:8106/hello -d 'who=world'
Hello world

Or the same with some JSON input:

$ curl -X POST http://localhost:8106/hello -d '{"who": "world"}' -H "Content-Type: application/json"
Hello world


As developers, we want to generate configuration from templates.

Many template-related tools already exist. As examples, we can use PasteScript or Chef, depending on the use case. But those tools, and the others we know, have drawbacks we'd like to get rid of:

  • PasteScript has a shell interface (+ Python interface). There, we'd like some REST API.
  • PasteScript requires templates to be installed as Python projects, via entry points. There we'd like something simpler, or more flexible.
  • Chef is quite good for provisioning. But when we just want to generate a file from a template, it's overkill. We'd like something small we can install as saas or on our local workstation for one shot use.

Right now, features we'd like are:

  • lightweight: deploy it easily on the web, on a saas platform, or on local workstation.
  • configurable and extensible: easy tuning of routes, available template engines, authentication...

And here is a typical architecture we are thinking of:

  • a diecutter server, exposes API(s).
  • a diecutter client, i.e. an IHM such as a web interface, or some shell client.
  • you and your favorite client, i.e. web browser or shell.

Then what about putting it together with daybed for schema validation?

Comments !