gander

gander is a higher-performance and lower-friction chat experience for data scientists in RStudio and Positron. The package brings ellmer chats into your project sessions, automatically incorporating relevant context and streaming their responses directly into your documents.

Why not just chat? In many ways, working with gander is just like using a chat interface online or via shinychat. The gander assistant will automatically find the context it needs, though:

• Code context: File contents from elsewhere in the project you’re working on (e.g. the lines in your source file).
• Environment context: Variables, allowing the assistant to locate the column names and types in data frames you’re working with, images linked to in your documents, and function definitions. This is what differentiates gander from many other assistants out there; gander can interface with your R session to describe your computational environment.

Important

For now, gander is particularly effective in .R files but struggles to provide appropriately formatted responses in .Qmd and related file formats.

Installation

You can install gander like so:

pak::pak("simonpcouch/gander")

Then, ensure that you have an ANTHROPIC_API_KEY environment variable set, and you’re ready to go. If you’d like to use an LLM other than Anthropic’s Claude 3.5 Sonnet—like OpenAI’s GPT-4o or a local ollama model—and then set the .gander_fn and .gander_args options in your .Rprofile, like options(.gander_fn = "chat_openai") to use OpenAI’s GPT-4o or options(.gander_fn = "chat_openai", .gander_args = list(model = "gpt-4o-mini")) to use their GPT-4o-mini model.

The gander assistant is interfaced with the via the gander addin. For easiest access, we recommend registering the gander addin to a keyboard shortcut.

In RStudio, navigate to Tools > Modify Keyboard Shortcuts > Search "gander"—we suggest Ctrl+Alt+G (or Ctrl+Cmd+G on macOS).

In Positron, you’ll need to open the command palette, run “Open Keyboard Shortcuts (JSON)”, and paste the following into your keybindings.json:

    {
        "key": "Ctrl+Cmd+G",
        "command": "workbench.action.executeCode.console",
        "when": "editorTextFocus",
        "args": {
            "langId": "r",
            "code": "gander::gander_addin()",
            "focus": true
        }
    }

The analogous keybinding on non-macOS is Ctrl+Alt+G. That said, change the "key" entry to any keybinding you wish!

Once those steps are completed, you’re ready to use the gander assistant with a keyboard shortcut.

Example

This screencast demonstrates using the gander addin to iterate on a plot:

After loading the stackoverflow data into my environment, I highlight stackoverflow and type a plain language request to plot the data. The LLM’s response—which provides a minimal solution and refers to correct column names—is streamed directly into my document and selected in whole. After visually inspecting, I run the code, and then retrigger the addin and type a followup request to refine the plot, doing so iteratively until I’m satisfied with my results.

To read more about using the addin, check out the Getting Started vignette with vignette("gander", package = "gander").