Skip to content
OPQAI.
Sourced intermediate / 💻 Coding Free tools

Set Up Local LLMs for R and Python with Ollama

Job to be done: Set up and use local Large Language Models (LLMs) with R and Python code

🇳🇬 Ways to use this in Nigeria

Ideas to get you started, adapt to your situation.

  • 9-5 employee

    As a data analyst, use a local LLM to automate the classification of sensitive internal reports in R, ensuring company data never leaves your secure network.

  • Student

    For your data science assignment, use a local LLM to generate Python code snippets for data cleaning tasks, practicing AI integration without relying on internet access or paid APIs.

  • Entrepreneur

    As an entrepreneur building a data analytics tool, use a local LLM to prototype a text summarization feature for user reports, keeping client data private and avoiding API costs during development.

What you’ll get

You will learn how to install and run a Large Language Model (LLM) directly on your computer using Ollama. This allows you to use AI models with R and Python code without sending your data to external servers. This approach is great for privacy and for users with limited internet access.

Tools you need

  • Ollama (free): A tool to download, install, and run LLMs on your computer.
  • ellmer (free): An R package to connect your R code to LLMs.
  • chatlas (free): A Python package to connect your Python code to LLMs.
  • R or Python: Free programming languages to write your code in.

Steps

  1. Install Ollama: Download and install Ollama on your computer. The website provides instructions for Mac, Windows, and Linux.
    • For Mac users: You can install Ollama using Homebrew. Open your Terminal application and run: 
      brew install --cask ollama
    • After installation, you can check if it’s working by running ollama --version in your Terminal. You should see the Ollama version number.
  2. Download an LLM: Once Ollama is installed, you need to download a language model to run. The article mentions Gemma 3 1B as an option for resource-limited devices, but notes newer models exist and Gemma 3 1B might be too small for complex coding tasks. To download a model, open your Terminal and use the ollama run command followed by the model name. For example, to download and run Gemma 3 1B: 
    ollama run gemma:3b
    You should see output indicating the model is downloading and then a prompt where you can start chatting with it. Type exit to leave the chat.
  3. Install the R package (ellmer): If you use R, open your R console or RStudio and run: 
    install.packages("remotes")
remotes::install_github("posit-dev/ellmer")
    You should see messages indicating the package is being installed. Once done, you can load it with library(ellmer).
  4. Install the Python package (chatlas): If you use Python, you can install chatlas using pip: 
    pip install chatlas
    You should see messages indicating the package is being installed. Once done, you can import it in your Python script with import chatlas.
  5. Connect to your local LLM: Now you can tell your R or Python code to use the LLM running via Ollama. The exact function to use is chat() in both packages. The article doesn’t specify the exact arguments for connecting to a local Ollama instance, but a common way is to specify the model name. A starting point for R would be: 
    library(ellmer)
ellmer::chat(model = "gemma:3b")
    And for Python: 
    import chatlas
chatlas.chat(model="gemma:3b")
    You should be able to interact with the LLM through your code, sending prompts and receiving responses that stay on your machine.

Original source

This guide is based on a blog post by Isabella Velásquez on the Posit blog, explaining how to set up and use local Large Language Models (LLMs) with R and Python. It introduces the free and open-source packages ellmer for R and chatlas for Python, and shows how to connect them to Ollama, a tool for running LLMs locally.

Notes & variations

  • Free tier alternative: Ollama, ellmer, and chatlas are all free and open-source, so there are no free tier limitations. The main cost is your computer’s processing power and storage.
  • Common mistake: Trying to run very large or complex LLMs on older or less powerful computers can lead to slow performance or errors. Start with smaller models like Gemma 3 1B if you have limited hardware.
  • Tip for better results: While local LLMs are great for privacy and cost, for the highest quality AI responses, consider paid services like Anthropic’s Claude Sonnet 4 or OpenAI’s GPT-4. You can often connect these paid services to ellmer and chatlas as well, if you have an account and API key.

Keep going

More Coding workflows