Installation and setup

Installation and setup is the step that puts the nemoguardrails package in your Python environment and a free Groq API key in your shell, which is everything this course needs to run.

Last updated: 30 Sep, 2026 · NeMo Guardrails 0.24.1

NeMo Guardrails does not ship a model. It calls one over HTTP, so you need a key for a hosted model. Groq's free plan runs openai/gpt-oss-120b, and every example in this course uses it.

Installing nemoguardrails

pip install "nemoguardrails==0.24.1"

The package brings the runtime, the Colang parser and a command line. It also installs FastEmbed, the library behind its default embedding model; this course configures a different search in the Embedding model lesson, so no model weights are downloaded.

Getting a Groq API key

As in the clip: sign in at console.groq.com/keys, create an API key and copy it. The free plan needs no card. It has a daily token limit per key, which is plenty for this course.

The video's demo app asks for two keys, a main one and a backup, and says the same key works in both places. The code in this course reads one variable, GROQ_API_KEY.

export GROQ_API_KEY=gsk_...

Checking the install

from importlib.metadata import version
import os

print(version("nemoguardrails"))
print("GROQ_API_KEY" in os.environ)

0.24.1
True

The first line is the installed version. True means the key is set in the shell that runs Python. A False here is the reason for most first-run errors, and the Main model and API key lesson shows the error it causes.

Keys this course uses

Where you install it

• In a fresh virtual environment or a Colab notebook, one per project.
• Next to an existing application, when the rails will sit in front of a chatbot or a RAG pipeline you already have.

Related

• Previous: NeMo Guardrails overview
• Next: An assistant that answers anything
• Reference: Installation guide