Upcoming Workshop: Agentic LLM Workflows

A hands-on workshop on using large language models as agents: how to frame a task, assess the result, work safely with tools, and choose an appropriate model. We will work in VS Code with GitHub Copilot and, where useful, connect personal Gemini and Mistral models.

The date, venue, and workshop-repository link will be shared with participants before the workshop. You can complete the preparation steps below in advance.

What to prepare

a laptop running a current version of Windows, macOS, or Linux with internet access,
a GitHub account for VS Code and GitHub Copilot,
VS Code installed and signed in with the same GitHub account,
optionally, your own Gemini and/or Mistral key, stored outside the repository.

API keys are not required to get started, but they let you try more providers and models during the workshop.

1. Create a GitHub account

Visit the GitHub sign-up page.
Create a personal account or continue with Google.
Verify your email address.
Enable two-factor authentication (2FA).
Sign in and, before the workshop, confirm that the workshop repository link provided to you opens.
Activate Copilot Free if it is available for your account.

GitHub offers Copilot Free with limited access to AI features, agents, and automatic model selection. Terms and allowances can change, so use the current plans page rather than relying on fixed quota numbers. GitHub also recommends enabling 2FA after account creation; see its account-creation instructions.

Readiness check: You are ready when you can sign in to GitHub, open the workshop repository, and see Copilot Chat in VS Code.

2. Install Visual Studio Code

Windows

Download the User Installer from the official VS Code download page.
Run the downloaded installer.
Keep the option to add VS Code to PATH enabled.
Start VS Code.
Sign in with the GitHub account prepared above.
Open the Chat or Agents view and activate Copilot.
Download or clone the workshop repository.
Open the repository as a folder, not as individual files.

Microsoft recommends User Setup for most Windows users because it does not require administrator rights and provides smoother updates. Follow the official Windows installation guide for the detailed steps.

Use the relevant official guide for macOS or Linux. If Git is new to you, use Code > Download ZIP in the workshop repository as a fallback. For the full clone, commit, and branch workflow, install Git.

VS Code with the Chat view beside the editor — The Chat view in VS Code sits beside the editor. Open it from the Chat menu, with Chat: Open Chat, or on Windows with Ctrl+Alt+I.

For more ways to open and arrange Chat and Agents, see the official VS Code Chat-view guide.

3. Install Git

Git is needed for the full workflow of cloning the repository, using branches, making commits, and sharing changes. If you only want to read the workshop material for now, use Code > Download ZIP in the repository; Git is recommended for the hands-on exercises.

Windows

Download the current installer from the official Git for Windows page.
Run the installer and keep the standard defaults, including the option to use Git from the command line.
When installation is complete, open a new PowerShell or Git Bash window.
Confirm the installation by running:

git --version

If you use Windows Package Manager, you can instead install Git with winget install --id Git.Git -e --source winget.

macOS and Linux

macOS: In Terminal, run xcode-select --install; alternatively, if you use Homebrew, run brew install git. See the official macOS installation options.
Debian/Ubuntu: sudo apt install git
Fedora: sudo dnf install git
Other distributions: use your distribution’s package manager as described in the official Linux guide.

On any operating system, confirm the installation again with git --version. If you will make your own commits, set your name and email once; this information becomes part of commit history:

git config --global user.name "Your name"
git config --global user.email "[email protected]"

See the Git first-time setup guide for details.

Readiness check: You are ready when git --version prints a version and VS Code shows Source Control after you open the repository.

4. Obtain a Gemini API key

Sign in to Google AI Studio.
Accept the applicable terms.
Open the Projects or API Keys section.
Use the automatically created project or import an existing Google Cloud project.
Create or display the Gemini API key.
Copy it into a password manager.
Do not place it inside the workshop repository.
In VS Code, run Chat: Manage Language Models, choose Add Models, select Gemini, and enter the key.
Select a Gemini model that supports tool calling.
Run the connection test supplied in the workshop repository.

Google provides a limited free Gemini Developer API tier for selected models. Submitted content on the free tier can be used to improve Google products, so do not use it with confidential workplace or personal data. Check the current conditions on Gemini API pricing.

Google recommends environment variables such as GEMINI_API_KEY and explicitly warns against committing keys to Git or exposing them in client-side applications. See the Gemini API-key guide. In VS Code, add personal models through Language Models; models must support tool calling to be available to agents. See the VS Code language-model documentation.

5. Obtain a Mistral API key

Create a Mistral Studio account.
Open the API Keys section.
Select Create new key.
Give the key a recognizable name, for example agentic-llm-workshop.
Set an expiration date if the interface offers one.
Copy it immediately; the complete value is shown only once.
Store it securely outside the repository.
Connect it through a built-in VS Code provider if available, a compatible provider extension, or a custom API endpoint.
Select a tool-calling model and run the connection test.

Mistral states that Free mode requires no credit card but is subject to usage and rate limits. Use the current official tutorial for activating an account and generating an API key. Provider and model availability in VS Code can change; if Mistral is not listed as a built-in provider, use a compatible extension or a custom endpoint as described in the VS Code documentation.

Safe use during the workshop

Never put keys, passwords, internal documents, or personal data into chat, prompts, or the repository.
Before running a tool, check what the agent proposes to do and which files it can change.
Review agent changes like ordinary code changes: inspect the diff and run tests.
If you are unsure, pause and ask during the workshop.

I look forward to experimenting with agents, tools, and models together.

Michael Princ

Tag: github-copilot