Support cookbook in connectapi: Creating Execution Environments, Viewing Execution Environment Runtimes

[toph-allen]

• Creating Execution Environments
• Viewing Execution Environment Runtimes

The code below imagines a basic / bare-bones approach, but execution environments are a resource class with CRUD methods, so — thinking about it, that should be our approach.

So, this looks like

Tasks

Beta Give feedback

1. Add a class for execution environments
2. Add CRUD functions for environments.

---

Viewing Execution Environment Runtimes

library(connectapi)

IMAGE_NAME <- "ghcr.io/rstudio/content-pro:r4.1.3-py3.10.4-bionic"

client <- connect()

all_images <- get_environments(client)
environments[[IMAGE_NAME]]$runtimes

# A tibble: 3 × 3
   language version path
   <chr>    <chr>   <chr>
 1 Python   3.10.4  /opt/python/3.10.4/bin/python3
 2 Quarto   1.0.35  /opt/quarto/bin/quarto
 3 R        4.1.3   /opt/R/4.1.3/bin/R

Notes

• The subsetting looks kinda janky. I'm assuming here that environments is a list of environments with more than just runtimes, but if all we care about here is runtimes, I guess I could imagine a few return formats: (1) a data frame with a column for environment name; (2) a list of environments, each a list with multiple fields including runtimes; (3) a list of environments, each an object with multiple fields including runtimes; (4) a list of runtime tibbles. [Update] We should do number (3).

Creating Execution Environments

library(connectapi)

CONTAINER_REGISTRY = "myorg/myrepo"

client = connect()

create_environment(
  title = "Custom Image Classifier",
  description = "My custom image classifier",
  cluster_name = "Kubernetes",
  name = paste0(CONTAINER_REGISTRY, "/image-classifier:jammy"),
  matching = "exact",
  r_version = "4.2.3",
  r_path = "/opt/R/4.2.3/bin/R",
  python_version = "3.9.14",
  python_path = "/opt/python/3.9.14/python"
)

A JSON object representing the environment is returned.

create_execution_environment(title = "Project alpha", description = , ...)

Accept many things, but at the very outset covert to a single type.

It's reasonable that if you know a version, you can guess the path. Like 4.2.3, guess the path?

What if it's a vector of versions?

Could just have just separate parameters for python, quarto, r, tensorflow. list(list(path = "opt/234/r", version = "234"), list(path = "opt/567/r", version = "567")).