Edit on GitHub

Contribute an Extension

This page explains how to contribute an open source extension. If you’re interested in only using extensions, visit Extensions.

Create and test a function in your Tiltfile

Tiltfiles are written in a Python dialect called Starlark. And so for the purpose of writing a new extension, it is no different from writing your Tiltfile, namely following typical Python syntax.

Create a new function, following the def func_name(args): syntax and add it to your Tiltfile. (Refer to any existing extension as an example.) Invoke the function later in your Tiltfile. Run Tilt as normal, and verify that the function works as expected.

Package your function and submit a pull request

Clone the tilt-extensions repo, and create a new extension, following the directory structure of other existing extensions. Namely, there should be a root-level directory with the name of your extension, and a Tiltfile inside that directory. Copy the function you previously tested into that Titfile. Also update README.md, explaining your extension. I.e. you should have these changes:

extension_name/Tiltfile
README.md

Create and submit a pull request to the repo, and @-mention @victorwuky for review.

Currently there’s no way to directly test the end-to-end workflow of using an extension. The Tilt team will ensure that the extension is working correctly before publishing it.

Next steps

If you run into any problems, contact us. If you have an extension idea (but aren’t interested in contributing), request it.

Example extension: API Server Logs

Depending on the kind of Kubernetes work you’re doing it can be valuable to see the Kubernetes API Server logs. For example, if you were iterating on a Kubernetes controller that interacts heavily with the Kubernetes API. By default Tilt doesn’t display those logs because they don’t come from a resource that Tilt deployed. However, it’s pretty easy to get tilt to display them using a local_resource:

local_resource('API Server Logs', '', serve_cmd='kubectl logs -f -n kube-system kube-apiserver-docker-desktop')

This only works for Docker for Desktop on macOS, since it’s hard coded to that pod name. But it would be easy to write a query to get the API server pod for any arbitrary Kubernetes cluster:

api_server_pod_name = str(local('kubectl get pods --namespace kube-system -o=jsonpath="{.items..metadata.name}" -l component=kube-apiserver')).rstrip(\n)

Then you could compose them together in a function:

def api_server_logs():
  api_server_pod_name = str(local('kubectl get pods --namespace kube-system -o=jsonpath="{.items..metadata.name}" -l component=kube-apiserver')).rstrip(\n)
  local_resource('API Server Logs', '', serve_cmd='kubectl logs -f -n %s' % api_server_pod_name )

Example extension: Jest Test Runner

Tilt is great at building your code and running your services, and with extensions it’s easy to make Tilt great at running your tests. Let’s use the Jest JavaScript test runner as an example.

To run Jest you simply do yarn run jest. Unlike a lot of test runners Jest runs in the foreground, watches your filesystem for changes and runs the correspond tests. So to translate yarn run jest in to a local_resource we’ll actually set it as the serve cmd:

local_resource("jest", "", serve_cmd="yarn run jest")

This works well except there’s no indication in the Tilt UI when your tests fail. That’s because Tilt only considers a local_resource as failing if the process exits, which Jest never does. Fortunately you can tell Jest to exit by passing the --bail flag. Let’s just wrap it in a function that takes a path to run Jest from and we have ourselves a Jest extension:

def jest(path):
  local_resource("jest", "", serve_cmd="cd %s && yarn run jest --bail" % path)

Now Tilt is running your tests, how cool is that?

 
 
 

Was this doc helpful?