02 — Setting up the environment + API key

Duration: 10 min Prerequisites: Python 3.9+, an OpenAI account (free to create).

1. Clone the two repos

mkdir project-2
cd project-2

# Shared library (mandatory)
git clone https://github.com/inskillflow/csv-llm-shared.git

# The OpenAI app
git clone https://github.com/inskillflow/csv-llm-openai.git

If you already did Project 1, you already have csv-llm-shared/ cloned — you can reuse it (just clone csv-llm-openai/ next to it).

2. Install the Python deps

pip install -r csv-llm-shared/requirements.txt    # pandas, plotly, pyyaml
pip install -r csv-llm-openai/requirements.txt    # + streamlit, openai, python-dotenv

Sanity check:

python -c "import pandas, plotly, yaml, streamlit, openai, dotenv; print('OK')"

3. Create an OpenAI API key

1. Sign in at platform.openai.com.
2. Go to Settings → Billing and add at least $1 of credit. (You can spend more later but $1 is the minimum payment.)
3. Go to Settings → API keys → Create new secret key.
4. Give it a name like csv-llm-openai-course and click Create.
5. Copy the key (sk-...). You won’t see it again. Save it somewhere temporary (a sticky note window, not committed anywhere).

Never commit your key

The OpenAI key gives full account access. If it leaks (in a public repo, in a screenshot, in a Discord paste), bots will find it within minutes and burn your credit. Always revoke a leaked key immediately at the keys page.

4. Store the key in .env

The repo ships with a template:

copy csv-llm-openai\.env.example csv-llm-openai\.env

Open csv-llm-openai\.env in your editor and paste:

OPENAI_API_KEY=sk-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

5. Verify .env is gitignored

csv-llm-openai/.gitignore already contains:

.env
.cache/
__pycache__/

Double-check with git status after copying:

cd csv-llm-openai
git status

Expected: no mention of .env. If you see .env listed as untracked, STOP, and verify .gitignore is correct before going further.

Belt and suspenders

You can also set the key as an OS-level environment variable instead of (or in addition to) .env. The code reads both. Set it system-wide once and you’re done forever:

# Windows PowerShell (persistent)
[Environment]::SetEnvironmentVariable("OPENAI_API_KEY", "sk-...", "User")

# macOS/Linux: add to ~/.zshrc or ~/.bashrc
export OPENAI_API_KEY="sk-..."

6. Smoke test (no Streamlit)

Quick sanity check that your key works:

python -c @"
import os
from dotenv import load_dotenv
load_dotenv('csv-llm-openai/.env')
from openai import OpenAI
client = OpenAI()
r = client.chat.completions.create(
    model='gpt-4o-mini',
    messages=[{'role':'user','content':'Say OK in one word.'}],
)
print(r.choices[0].message.content)
"@

Expected output:

OK

If you see OK, you’re good. If you see AuthenticationError, re-check the key and that the .env was loaded.

7. Verify the data path

The app reads csv-llm-openai/data/data1-anonymized.csv by default. Confirm it’s there:

ls csv-llm-openai\data\data1-anonymized.csv
# expected: 1 file, ~57 KB

It should exist (the repo ships with it).

Cost expectations for the course

Across the 7 chapters and 4 canonical questions, you’ll make roughly:

• 10 chat completions (test calls, demos)
• ~6,000 input tokens + ~3,000 output tokens total
• Total cost: under $0.005 — half a cent.

You’ll hit your $1 minimum hundreds of courses before running out.

Troubleshooting

Error	Likely cause	Fix
AuthenticationError: Incorrect API key	Wrong key, or trailing whitespace in .env	Re-copy the key, no quotes, no spaces
RateLimitError: 429	New account, billing not active	Add $1 of credit in Settings → Billing
ModuleNotFoundError: openai	Pip install missed	pip install -r csv-llm-openai/requirements.txt
.env shows up in git status	Wrong .gitignore	Add .env to csv-llm-openai/.gitignore

Takeaways

• One key, one .env file, one .gitignore entry — that’s the whole secret-management story.
• Smoke test with openai.chat.completions.create() confirms the key works.
• Cost for the full course: under one cent.

Next: llm.py side-by-side — Ollama vs OpenAI →