Generative AI for Beginners

What it is

Generative AI for Beginners is a free course from Microsoft Cloud Advocates, kept as a GitHub repository under the MIT license. Its README says the course has 21 lessons and that each one covers its own topic, so you can start where you like. A “Learn” lesson explains a concept. A “Build” lesson explains a concept and adds code in Python and TypeScript where possible. The README says each lesson includes a short video, a written lesson and links to extra learning, though its table marks a few videos as coming soon.

1. Fork the repository to your GitHub account
2. Open it in a Codespace, or clone it to your computer
3. Copy .env.copy to .env and fill in your model provider
4. Read a lesson in its README
5. Run its Python or TypeScript sample

Who it suits

Developers who are new to generative AI and know a little Python or TypeScript. The README calls that knowledge helpful and links a Python course and a TypeScript course for absolute beginners. This is a teaching repository, and AGENTS.md says its examples are intentionally simple and not production code.

Skip it if you want a finished app to deploy. The README points to a separate collection of code samples for more advanced work, and to a separate .NET edition.

Where to use it

The course is Markdown lessons, Python and TypeScript samples and Jupyter notebooks, so it runs wherever those do. The Course Setup page recommends a GitHub Codespace.

• GitHub Codespaces: a browser-based VS Code with Python 3, Node.js, .NET and Java already inside
• Your own computer: a Python virtual environment, Miniconda, or classic Jupyter in your browser
• VS Code with Docker Desktop: the Dev Container, the same environment a Codespace uses
• Your browser with no setup: each coding lesson’s README shows its code and outputs

The code reaches a model through a provider, and an assignment that needs a specific provider carries its tag in its filename.

How to set it up

Get it saves a zip of Mindket’s setup notes in your downloads folder, with each path’s steps from Microsoft’s setup pages. The main route is a fork opened in a Codespace, or a local clone with a Python virtual environment.

1. Fork the repository to your GitHub account.
2. Open your fork in a Codespace. On the fork, pick Code, Codespaces, Create codespace on main. Or clone it and make a virtual environment, as below.
3. Run pip install -r requirements.txt. It installs the Python packages. AGENTS.md says a Dev Container does this for you, though the post-create script it names installs only python-dotenv, openai and the lint and test tools.
4. Copy .env.copy to .env and fill it in. Use the values for the provider you picked.
5. Open a lesson’s README and run its sample.

git clone https://github.com/<your-github>/generative-ai-for-beginners
cd generative-ai-for-beginners
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.copy .env

On Windows PowerShell, the Local Setup page activates the environment with this line instead of the source line.

.\.venv\Scripts\activate

The repository holds many translations, which make a clone large. The README gives a sparse checkout that leaves them out.

git clone --filter=blob:none --sparse https://github.com/microsoft/generative-ai-for-beginners.git
cd generative-ai-for-beginners
git sparse-checkout set --no-cone '/*' '!translations' '!translated_images'

How to use it

Start anywhere. Read a Learn lesson’s README. For a Build lesson, open its folder and run the sample for the provider you set up. AGENTS.md gives this pattern for a Python sample:

cd 06-text-generation-apps/python
python aoai-app.py

The oai files use the OpenAI API, the aoai files use Azure OpenAI and the githubmodels files use Foundry Models. For Lesson 06 in TypeScript, AGENTS.md gives these lines:

cd 06-text-generation-apps/typescript/recipe-app
npm install
npm run build
npm start

Notebooks open in Jupyter or VS Code. The Local Setup page’s example is 08-building-search-applications/python/oai-solution.ipynb.

Example

You want to try Lesson 06 on your own OpenAI account. You open a Codespace from your fork and save OPENAI_API_KEY as a Codespaces secret, which the Cloud Setup page says the code picks up. In its terminal you run:

cd 06-text-generation-apps/python
python oai-app.py

The script sends the prompt “Complete the following: Once upon a time there was a” to the gpt-5-mini model and prints the reply. To see how the wording steers the answer, open the file, change the prompt and run it again.

For newcomers and experts

If you are new. A large language model, or LLM, is the kind of model behind chat tools. Your code sends it text through an API and gets text back. Lesson 01 explains what generative AI is and how LLMs work, and Lesson 02 helps you choose a model. Start with Lesson 01. You can read before you set up any account, because each coding lesson’s README shows its code and outputs.

For experts. AGENTS.md sets the model rules. The default chat model is gpt-5-mini, a reasoning model that rejects temperature and top_p and takes max_output_tokens or max_completion_tokens in place of max_tokens. The temperature demos use a Llama model through the Foundry Models endpoint, named by AZURE_INFERENCE_CHAT_MODEL.

Lesson 18 keeps gpt-4.1-mini because GPT-5 only supports reinforcement fine-tuning. Lessons 20 and 21 keep temperature and max_tokens because they target Mistral and Llama models. In TypeScript, the openai package serves Azure OpenAI through the /openai/v1/ endpoint and the Responses API, and @azure-rest/ai-inference serves Foundry Models.

The Providers page says Foundry Local exposes an OpenAI-compatible endpoint, so you can reuse most of the sample code with minimal changes.

Good to know

The lessons and code sit on your machine or in your Codespace, but every model call goes to the cloud provider you configure unless you run Foundry Local or Ollama.

• Accounts and cost: you use your own provider accounts, and each cloud provider has its own pricing, which the Providers page links. Related assignments error out on missing credentials. Foundry Local runs on your device and needs no key
• Your keys: keep them in .env, which the Local Setup page says is already in .gitignore, or in Codespaces secrets. Never commit them
• Your prompts: every call goes to the provider you configure. Foundry Local and Ollama keep them on your device
• GitHub Models: the docs say it retires at the end of July 2026, and that date has passed. Files tagged githubmodels use Foundry Models credentials
• Setup pages disagree: AGENTS.md says Python 3.9 or newer, the Local Setup page says 3.10 or newer, and the .python-version file reads 3.12.10. The pinned numpy and pandas need 3.11 or newer on PyPI. The Course Setup page says Azure OpenAI access is open with no application, while the Providers table says “Must Apply Ahead For Access”
• Gaps in the docs: after its troubleshooting table, the Course Setup page shows the end of the .env guide without its heading or first step. That end is the touch and echo commands, then steps 3 to 6. Use the Local Setup page for that guide. Local Setup says to make sure Jupyter is installed and gives no command for it
• Tracking links: the repository’s rules put a tracking ID on links to Microsoft domains, so a click from a lesson carries it
• Removal: the pages give no uninstall steps. Delete the folder, or stop or delete the codespace
• License: MIT. The LICENSE file reads “Copyright (c) Microsoft Corporation.”