Last updated on 2026-02-12 | Edit this page

Session 2

Required Self-Study

Session Plan

Autocompletion

  • Shortcut to switch Copilot on/off
    • via “Command Palette”:
      • SHIFT+CTRL+P or SHIFT+COMMAND+P > type “Copilot”
    • setup of personal key combination
      • Tools > Modify Keyboard Shortcuts..
      • use filter field with “Copilot” to find
        • Copilot: Toggle Automatic Completion
      • click into center column of the table to set your key combination
        • e.g. SHIFT+CTRL+ALT+P or whatever you like
  • Coding examples … to experience context dependency (comments)
    • enable Copilot and autocompletion
    • create two new R scripts
      • copy & paste the following code snippets into the two scripts
    • try to complete the code by accepting Copilot suggestions
      • compare the results!

R 

library(tidyverse)
data("storms")

# visualize for each year the maximal wind speed and the number of storms
# in a bar chart with two bars per year

R 

library(tidyverse)
data("storms")

# identify for each storm in each year the maximal wind speed
# visualize for each year the maximal wind speed of the strongest storm
# and the number of storms in a bar chart with two bars per year
# ensure that the year names on the x-axis are readable

  • Discuss within your group:

    • ? do you see differences?
    • ? which code seems more appropriate/correct?
    • ? do you understand all code elements?

  • ! Context matters …

  • ? What happened to enable the code suggestions?

    • Copilot uses the context of the current file
    • ! requires upload of your code!
    • ? drawbacks? (security, online traffic, energy consumption, bubble, …)
      • ! use with care!

  • Setup of Copilot options

    • Tools > Global Options
    • “Copilot” menu on the left
      • for most use cases: sufficient to disable indexing of whole project!

Context Definition

  • so far: context via comments
    • PRO: simple and quick to use
    • CON: general requirements and coding constraints need to be defined in each document
  • alternative: context via AGENTS.md files
    • PRO: central definition of requirements and constraints for the whole project
    • CON: requires some setup and maintenance = additional work but WORTH IT!
    • typically requires code organisation in terms of projects (RStudio Projects)
    • IMPORTANT: need to activate “Index project files with GitHub Copilot” in RStudio > Copilot options!
  • ? Who is familiar with RStudio Projects?
    • idea and setup
      • relation of project and working directory
      • relation of project and git repository
    • switching projects
    • AGENTS.md file
      • project root directory vs. subdirectory
      • verbose description of coding requirements and constraints
        • even more detailed than comments in code files
        • use examples
        • update regularly while working on the project
  • Exercise: Setup of RStudio Project with AGENTS.md file
    • create a new RStudio Project
    • add a new R script with the short description from above
    • create an AGENTS.md file in the project root directory
      • provide detailed instructions for AI on general details on
        • coding style (e.g. where to put documenting comments)
        • visualization style (colors, fonts, themes, …)
        • image generation (output file format, sizes, resolution, …)
        • any other requirements or constraints you find important
      • ! you might want to use AI-supported autocompletion for this…!
    • reopen the created R script with the incomplete code
      • try to complete the code by accepting Copilot suggestions
    • compare the results with those from the previous exercise
    • double check your AGENTS.md file
      • ? did Copilot follow your instructions?
      • ? what could be improved in your AGENTS.md file?
        • too specific vs. too general?