Skip to content

setup.md

Latest commit

 

History

History
81 lines (56 loc) · 2.02 KB

setup.md

File metadata and controls

81 lines (56 loc) · 2.02 KB
Raw

Setup — about 10 minutes

You need: a laptop, Python 3.10 or newer, and an internet connection. You do not need a GPU, a cloud account, or a credit card.

1. Get a free API key (~2 minutes)

The tutorial calls a hosted language model. The default is NVIDIA's free tier — generous enough for the whole tutorial.

  1. Go to https://build.nvidia.com and sign in (a free account).
  2. Pick any model (for example NVIDIA Nemotron™).
  3. Click Get API Key. Copy the key — it starts with nvapi-.

Prefer another provider? Any OpenAI-compatible endpoint works — OpenAI, a local Ollama, or a self-hosted vLLM. See step 3.

2. Install the dependencies (~5 minutes)

Clone or download this repository, then from its root:

# Recommended: uv (https://docs.astral.sh/uv/)
uv sync

# Or with plain pip + a virtual environment
python -m venv .venv && source .venv/bin/activate
pip install -e .

3. Configure your key (~1 minute)

cp .env.example .env

Open .env and paste your key into LLM_API_KEY and EMBED_API_KEY.

To use a different provider, change only the three LLM_* lines — for example, for a local Ollama:

LLM_BASE_URL=http://localhost:11434/v1
LLM_API_KEY=ollama
LLM_MODEL=llama3.1

4. Check the connection

python llm.py

Expected output:

Endpoint : https://integrate.api.nvidia.com/v1
Model    : nvidia/nemotron-3-nano-30b-a3b
Calling the model ...
Reply    : connection ok

If you see LLM_API_KEY is not set, your .env still has the placeholder.

5. Open the notebooks

jupyter lab

Start with part-1-build-an-agent/01-raw-agent.ipynb. Run the cells top to bottom.

A note on the free tier

The free tier is rate-limited (roughly 40 requests per minute) and metered in credits. Parts 1 and 2 use very few calls. Part 3 evaluates a small dataset and is deliberately kept to ~25 items so it stays inside the free allowance. If you hit a rate limit, wait a minute and re-run the cell.