Azure AI Foundry Tutorials - Hello World Example

AI agents are basically automated systems that can understand what's happening around them, make decisions, and take actions to achieve specific goals. Before we look at how to create AI agents using azure we will first need to learn azure foundry.

What is Azure AI Foundry?

Before we look at what is Azure AI foundry, let us first look at its history.

Platform	Key Milestones & Dates	Notes / Purpose
Azure OpenAI Service	- Launched (preview) in November 2021. Sources: Microsoft Azure (+2), TechRepublic (+2) - General Availability announced 16 January 2023. Sources: Microsoft Azure (+1)	Provides access to large language models (LLMs) such as GPT-3.5, Codex, and DALL-E 2 on Azure infrastructure. It's the foundational model-serving layer for generative AI on Azure.
Azure AI Studio	- Initial announcement/preview: 15 November 2023 (approx.) Sources: InfoWorld (+1) - General Availability declared 21 May 2024. Sources: Visual Studio Magazine	A development environment (studio) for building AI applications - model selection, grounding, retrieval-augmented generation (RAG), and agent workflows. It sits on top of the model services like Azure OpenAI Service.
Azure AI Foundry	- Announced at Microsoft Ignite 2024 on 19 November 2024. Sources: TechCommunity.Microsoft.com (+1) - Rebranding of Studio into Foundry and unified platform context. Sources: InfoWorld (+2), Microsoft Azure (+2)	A unified platform for enterprise-scale agentic AI - combining models, agents, tooling, governance, and monitoring. It subsumes Studio functionality and integrates OpenAI Service, speech, vision, language, and agent management under one umbrella.

Azure AI Foundry is a unified platform for building, customizing, and managing generative AI applications and intelligent agents. Formerly known as Azure AI Studio, it provides a comprehensive set of tools, models, and capabilities in a single environment to help developers accelerate the creation and deployment of AI solutions.

• Accessed from Separate Domain: ai.azure.com
• Model Catalog: Offers access to foundational, open-source, task-specific, and industry-specific AI models.
• Integration with Developer Tools: Works with Copilot Studio, Visual Studio, GitHub, and the Azure AI Foundry SDK.
• Core AI Services: Includes Azure OpenAI Service, AI Search, AI Agent Service, and AI Content Safety.
• Customization & Governance: Supports model fine-tuning, evaluation, compliance, and responsible AI practices.

Azure AI Foundry - Hub and Project

Azure AI Hub is the main workshop or head office for all our company's AI efforts. Whereas a Project can be thought of as a separate, focused workspace inside the main Hub workshop. It's where a smaller team works on one specific AI goal.

Hubs

• Acts as a central AI governance and management layer across multiple projects.
• Facilitates sharing and reuse of AI assets (models, datasets, prompts) across teams.
• Best for organizations managing multiple AI projects with shared resources.

Projects

• Designed for developing and managing AI models and workflows.
• A workspace for experimenting, fine-tuning, and deploying AI models.
• Supports integration with various AI services, data sources, and pipelines.
• Ideal for a single AI development lifecycle, from experimentation to production.

Implementation

Go to azure portal - Azure Portal. Search Azure Foundry and select it.

We will first be creating ai hub

In the ai foundry portal create a new ai hub named foundry-hello-world.





Once we create a new hub we can see that azure also created related resources -

• Machine Learning Resources - This is the Azure AI / ML workspace. It is the central hub for all our machine learning activities.
• Azure Cognitive Services - Using it we get access to pre-built, high-level AI capabilities from Microsoft, such as Vision (e.g., image analysis, face recognition), Speech (e.g., speech-to-text, text-to-speech). Language (e.g., sentiment analysis, translation). The Azure AI Hub often uses this account for features that leverage these pre-trained models.
• Azure Keyvault - This is where secrets are stored securely
• Azure Storage - Used to store data and artifacts

We go to the created hub resource.

Currently there is nothing in connected resources. Next we will create a new project named new-project inside the created ai hub.

We can now see new connected resources are created.

• Storage Account - Workspace artifact - Store artifacts like context window text for a prompt.
• Storage Account - Workspace blob storage - Store data uploaded by user like images, documents etc.

Inside this ai project we can deploy new base model. We will be deploying the o3-mini model.



Once created we can go to the playground to test it.

Also if we go to the connected resource we can see an additional resource which is a keyvault api key. This key is used by the ai hub connect to the deployed model.