Tech Requirements

Six pieces, one toolchain

You will install a Python stack, a code editor, a terminal AI agent, and the keys that let them talk to a model. Here is how the pieces fit together.

Two ground rules

• Use the official downloads only. Every link on these slides points to the vendor's own site. Avoid mirrors and "download helper" sites.
• Verify each step. After every install there is a one-line check (a --version command). If the check fails, fix it before moving on — later steps depend on earlier ones.

Download and run the installer for your OS

Get the installer from anaconda.com/download. Pick the build that matches your machine.

bash ~/Downloads/Anaconda3-2024.10-1-Linux-x86_64.sh

conda --version
python --version

What a conda environment is — and why you need one

An environment is an isolated, named Python installation with its own set of packages. Different projects need different — often conflicting — library versions; environments keep each project's dependencies in their own sealed box so they never clash.

conda activate llmfin        # switch this terminal INTO the env; prompt shows (llmfin)
python gen_king_analogy.py   # ...now plain python uses the env's interpreter

conda run -n llmfin python gen_king_analogy.py   # run ONE command in the env, no activation

• Why not just python? Bare python runs whichever interpreter is first on your PATH — usually base or the system Python, which lack the course libraries. You'd hit ModuleNotFoundError: No module named 'gensim'. Activating (or conda run) guarantees the right interpreter.
• Which env am I in? Run conda env list — the * marks the active one — or check the (llmfin) tag at the start of your prompt.

One environment, all the dependencies

Create an isolated environment named llmfin, then install the course requirements into it. Run these from the course folder (next section shows how to get it).

conda create -n llmfin python=3.11
conda activate llmfin
pip install -r code/requirements.txt
pip install -e code/

• requirements.txt pulls numpy, pandas, matplotlib, scikit-learn, jupyter, gensim, yfinance and more.
• pip install -e code/ installs the course's own llmfin helper package in editable mode.

python -m ipykernel install --user --name llmfin --display-name "Python (llmfin)"

Get the editor from code.visualstudio.com

winget install Microsoft.VisualStudioCode

brew install --cask visual-studio-code

sudo apt install ./code_*.deb   # Debian/Ubuntu
sudo dnf install ./code-*.rpm   # Fedora/RHEL

Four extensions — install from the Extensions view

Open the Extensions view (Ctrl+Shift+X, or Cmd+Shift+X on macOS), type each name in the search box, and click Install. The exact marketplace IDs:

code --install-extension ms-python.python
code --install-extension ms-toolsai.jupyter
code --install-extension anthropic.claude-code
code --install-extension saoudrizwan.claude-dev

Install the current LTS (v24)

Get the LTS build from nodejs.org/en/download. The npm route for Claude Code needs Node 18 or later — LTS is well above that.

brew install node

node --version
npm --version

The native installer is the recommended route

curl -fsSL https://claude.ai/install.sh | bash

irm https://claude.ai/install.ps1 | iex

npm install -g @anthropic-ai/claude-code

claude --version
claude doctor

Start it and log in

Open a terminal inside the project folder you want to work on and run:

claude

• On first run it opens your browser to log in and authenticate.
• Claude Code requires a Pro, Max, Team, Enterprise, or Console account — the free Claude.ai plan does not include it. (It can also run against Bedrock / Vertex / Foundry.)
• System requirements: macOS 13+, Windows 10 (1809+), or Ubuntu 20.04+ / Debian 10+, 4 GB+ RAM.

Install the extension, then connect a model

You already installed Cline (saoudrizwan.claude-dev) in §02. It is an autonomous agent that can create/edit files and run commands — with your approval at each step.

• Open Cline from the activity bar icon, or run Cline: Open In New Tab from the Command Palette.
• On first open, complete provider setup in Cline settings. You can use the built-in Cline provider (pay-as-you-go), or bring your own key — e.g. an Anthropic API key or an OpenRouter key.
• Pick a model (Claude works well) and you are ready.

What an API key actually does

Your code (or Cline) sends a request plus your secret key; the provider checks the key, runs the model, and bills your account.

Generate a key, then set it as an environment variable

1. Sign in at console.anthropic.com and add billing credits.
2. Go to Settings → API keys → Create Key. Copy it now — it looks like sk-ant-api03-... and is shown only once.
3. Store it as an environment variable named ANTHROPIC_API_KEY (the Anthropic SDK and tools read this automatically):

export ANTHROPIC_API_KEY="sk-ant-api03-..."

setx ANTHROPIC_API_KEY "sk-ant-api03-..."

pip install anthropic

Keep keys secret; two more keys for a few examples

Git — you need it to clone the course repository

Git is also what Claude Code uses for its Bash tool on Windows. Get it from git-scm.com/downloads, or via a package manager:

winget install Git.Git

brew install git

sudo apt install git   # Debian/Ubuntu
sudo dnf install git   # Fedora/RHEL

git --version
git config --global user.name  "Your Name"
git config --global user.email "you@example.com"

gh — GitHub from the terminal: auth, clone, pull requests, issues

The GitHub CLI logs you in to GitHub once and then handles cloning over HTTPS, pull requests, and issues. Get it from cli.github.com, or:

winget install GitHub.cli

brew install gh

sudo dnf install gh   # Fedora/RHEL

gh --version
gh auth login        # opens a browser to authenticate

Clone it and open it in VS Code

git clone https://github.com/jfimbett/llm-finance-book.git
cd llm-finance-book
conda activate llmfin
pip install -r code/requirements.txt
pip install -e code/
code .

Then open code/practicals/01-intro/practical.ipynb and select the Python (llmfin) kernel.

Run these — every line should succeed

conda --version
python --version
node --version
claude --version
git --version
gh --version
python -c "import pandas, numpy, sklearn, gensim; print('libs ok')"
python -c "import anthropic; print('sdk ok')"

Where to get unstuck

• Claude Code: run claude doctor; see the official troubleshooting guide at code.claude.com/docs.
• conda / Python: make sure you opened a new terminal after install and ran conda activate llmfin.
• VS Code can't find Python: re-run Python: Select Interpreter and choose the llmfin environment.
• Key not found: open a new terminal after setx / editing your shell profile, and confirm the variable is set.