Hi! A quick question about a detail which I just can't find in the docs (I suppose I'm missing something). I have a task which depends on a certain tool being installed on the system, i.e. Python is calling it using

subprocess

-- how can I ensure that the tool is installed and available to the task? Thank you! 😊

if your flows are being run in docker containers, you can write a custom dockerfile that installs that tool on top of the prefect base image, build that as an image, and refer to the image by name and tag in your deployment

and what happens if the tasks are supposed to be run externally, e.g. via Dask/Ray, or in some other distributed way? As I understand, the deployment dockerfile applies to the flow; how are the tasks managed? 🤔

if you're running an external cluster for distributed execution, I believe you'll have to similarly manage those dependencies separately

is there an example somewhere?

not that I'm aware of. starting up and managing dependencies for an external dask cluster might be a bit outside the scope of prefect-related knowledge

no worries, thanks! this will be a good starting point, but I have quite a bit of reading to do 😄