Create a lightweight chat client with the Foundry SDK
A lightweight chat client is a small application, often a console app or simple web app, that:
Connects to a Foundry project
Uses a deployed model
Sends a user message
Receives a generated response
Displays the response to the user
Microsoft’s Foundry quickstart shows this pattern using the Microsoft Foundry SDK, with support for languages such as Python, C#, TypeScript, Java, REST API, and the Foundry portal.
Exam-level workflow
1. Create or use a Foundry project
A Foundry project is the workspace where your AI solution is managed.
2. Deploy a model
For a chat client, you typically deploy a generative AI model such as a GPT-style model.
3. Install/use the Foundry SDK
For Python, Microsoft documents the package:
pip install azure-ai-projects
Microsoft’s current quickstart notes that the newer SDK uses Azure AI Projects 2.x and is not compatible with the older 1.x project SDK.
4. Authenticate
The app needs permission to access the Foundry project, commonly using Azure authentication such as Azure CLI sign-in or managed identity.
5. Send a chat request
The client sends messages such as:
System: You are a helpful assistant.
User: Summarise the benefits of Microsoft Fabric.
6. Display the response
The app receives the model output and shows it in the client interface.
Simple AI-901 explanation
Think of it like this:
User message
↓
Lightweight chat app
↓
Foundry SDK
↓
Deployed model in Foundry
↓
Generated response
↓
Displayed to user
What the exam is likely testing
Scenario clue Correct idea
“Build a small app that talks to a model” Use Foundry SDK
“Model must already be callable by code” Deploy the model first
“Need to test prompts in code” Send chat messages through SDK
“Need app integration” Use endpoint/SDK rather than only playground
“Human testing in portal” Use Foundry playground
Key distinction
The Foundry portal/playground is mainly for testing manually.
The Foundry SDK is for building an application that interacts with the model programmatically. Microsoft describes Foundry as a unified Azure platform for AI operations, model building, and application development.
Summary
For AI-901, remember:
Deploy model → connect with SDK → send prompt → receive response → display in chat client.