Setup
R + Python
- Clone this repository: https://github.com/chendaniely/2025-06-05-llm
git clone https://github.com/chendaniely/2025-06-05-llm.git
git clone git@github.com:chendaniely/2025-06-05-llm.git
- Use the provided
requirements.txt
andrenv.lock
file to create the virtual environments- R: you can use the
setup-r
Makefile
target - Python: if you have
uv
installed (https://docs.astral.sh/uv/) you can also use thesetup-python
target - There is a convenience
make setup
target that combines both for you
- R: you can use the
make setup
IDE
I’m using Positron: https://positron.posit.co/, but feel free to use VS Code and/or RStudio
GitHub Models
You will need to create a GitHub Personal Access Token (PAT). It does not need any context (e.g., repo, workflow, etc).
General instructions from the GitHub docs on creating a PAT: https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens#creating-a-personal-access-token-classic
Instructions from the GitHub Models docs: https://github.com/Azure-Samples/python-ai-agent-frameworks-demos/tree/main?tab=readme-ov-file#configuring-github-models
Local LLM: Ollama
- Download Ollama: https://ollama.com/
- Pick one of the many llama models on their model page from: https://ollama.com/search.
- UBC has a restriction on DeekSeek
- Pick any random model that will fit on your computer
- You can pick multiple models if you’d like, we will compare results during workshop.
- Here are a few example models with their download sizes you can try:
Model | Download Size | URL | Install Command |
---|---|---|---|
qwen3:0.6b | 523MB | https://ollama.com/library/qwen3 | ollama run qwen3:0.6b |
qwen | 5.2GB | - | ollama run qwen3 |
Phi 4 mini | 3.2GB | https://ollama.com/library/phi4-reasoning | ollama run phi4-mini-reasoning |
devstral | 14GB | https://ollama.com/library/devstral | ollama run devstral |
llama4 | 67GB | https://ollama.com/library/llama4 | ollama run llama4 |
llama4:128x17b | 245GB | - | ollama run llama4:128x17b |
(Optional): Chat provider with API
If you pay for Claude, OpenAI, etc access with their web/desktop application, this is a separate purchase for the API key. Depending on your usage, you may even find that paying for the API key could be cheaper!
Anthropic Claude
- Sign up at https://console.anthropic.com.
- Load up enough credit so you won’t be sad if something goes wrong.
- Create a key at https://console.anthropic.com/settings/keys
Google Gemini
- Log in to https://aistudio.google.com with a google account
- Click create API key & copy it to the clipboard.