Queries and chats also can embody uploaded photographs with the photographs argument.

ollamar

The ollamar bundle begins up equally, with a test_connection() operate to examine that R can connect with a operating Ollama server, and pull("the_model_name") to obtain the mannequin corresponding to pull("gemma3:4b") or pull("gemma3:12b").

The generate() operate generates one completion from an LLM and returns an httr2_response, which might then be processed by the resp_process() operate.

library(ollamar)

resp <- generate("gemma2", "What's ggplot2?")
resp_text <- resp_process(resp)

Or, you’ll be able to request a textual content response straight with a syntax corresponding to resp <- generate("gemma2", "What's ggplot2?", output = "textual content"). There’s an choice to stream the textual content with stream = TRUE:

resp <- generate("gemma2", "Inform me in regards to the information.desk R bundle", output = "textual content", stream = TRUE)

ollamar has different performance, together with producing textual content embeddings, defining and calling instruments, and requesting formatted JSON output. See particulars on GitHub.

rollama was created by Johannes B. Gruber; ollamar by by Hause Lin.

Roll your personal

If all you need is a primary chatbot interface for Ollama, one straightforward possibility is combining ellmer, shiny, and the shinychat bundle to make a easy Shiny app. As soon as these are put in, assuming you even have Ollama put in and operating, you’ll be able to run a primary script like this one:

library(shiny)
library(shinychat)

ui <- bslib::page_fluid(
  chat_ui("chat")
)

server <- operate(enter, output, session) {
  chat <- ellmer::chat_ollama(system_prompt = "You're a useful assistant", mannequin = "phi4")
  
  observeEvent(enter$chat_user_input, {
    stream <- chat$stream_async(enter$chat_user_input)
    chat_append("chat", stream)
  })
}

shinyApp(ui, server)

That ought to open a particularly primary chat interface with a mannequin hardcoded. For those who don’t decide a mannequin, the app received’t run. You’ll get an error message with the instruction to specify a mannequin together with these you’ve already put in domestically.

I’ve constructed a barely extra sturdy model of this, together with dropdown mannequin choice and a button to obtain the chat. You’ll be able to see that code here.

Conclusion

There are a rising variety of choices for utilizing giant language fashions with R, whether or not you wish to add performance to your scripts and apps, get assist together with your code, or run LLMs domestically with ollama. It’s price attempting a few choices in your use case to search out one that most closely fits each your wants and preferences.