AI-103 - Azure AI Apps and Agents Developer Associate

This lecture introduces the basic developer pattern for creating an AI agent using the OpenAI SDK.

In this hands-on lab, you will build your first AI agent using the OpenAI SDK. This gives you a practical starting point for understanding how agents receive user input, follow instructions, and return responses through code.

This lecture gives an overview of the Microsoft Foundry Agent Service. You will learn how Foundry helps developers create, configure, test, and manage agents that can use models, tools, knowledge, and instructions in a structured way.

In this lab, you will create a prompt agent directly in the Microsoft Foundry portal. This helps you understand the no-code and low-code experience for defining an agent, adding instructions, testing responses, and preparing the agent for real use.

This lecture explains what happens after an agent is created. You will learn how publishing makes an agent available, how routines can support repeatable agent behavior, and how distribution options help make the agent usable in business scenarios.

In this lab, you will build a prompt agent programmatically. This is important for developers because it shows how agent creation can be automated, version-controlled, and integrated into a repeatable development workflow.

Let's resolve our issue and try to register the agent again

In this lab, you will invoke the agent from your application code. You will see how to send a user message to the agent, receive the response, and understand the basic runtime flow between your application, the agent, and the deployed model.

This lecture introduces Microsoft Entra ID and explains why identity is important when building AI applications on Azure. You will learn how authentication and authorization help protect resources such as Foundry projects, models, storage accounts, and search services.

In this lecture, we take a quick tour of users and role-based access control in Microsoft Entra ID. You will see how permissions are assigned and why RBAC matters when giving users or applications access to Azure AI resources.

This lecture introduces tools in Microsoft Foundry agents. You will learn why tools are essential for agentic AI, allowing agents to search, calculate, call functions, analyze files, retrieve information, and interact with external systems instead of relying only on model knowledge.

In this lab, you will add and use the Web Search tool with an agent. This shows how an agent can access current information from the web, making responses more useful when questions depend on recent or external data.

In this lab, you will define an agent that can use custom function tools. You will learn how to describe the available functions, define the required parameters, and help the agent decide when a function should be called.

In this lab, you will complete the function calling workflow from the application side. You will see how your code handles the tool call, runs the appropriate business logic, returns the result to the agent, and allows the agent to produce the final answer.

In this lecture, we introduce the idea of hosted agents in Microsoft Foundry.

This lecture introduces Retrieval Augmented Generation, commonly known as RAG. You will learn why RAG is important when an AI application needs to answer questions using private, business-specific, or frequently changing information instead of relying only on the model’s built-in knowledge.

In this lab overview, we walk through the complete RAG scenario before building it step by step. You will understand the main flow: prepare documents, index them in Azure AI Search, retrieve relevant content, and use that content to ground the model’s final response.

In this lab, you will prepare the source data that will be used for the RAG solution. This includes organizing the knowledge files and making sure the data is ready to be uploaded, indexed, searched, and later used as grounding information for the AI application.

In this lab, we introduce Azure AI Search as the retrieval layer for the RAG solution. You will see how Azure AI Search helps store, index, and search your content so the application can find the most relevant information for a user question.

In this lab, you will build the search index that sits at the heart of the RAG solution. The index defines how the content is stored, what fields are searchable, and how the application can retrieve matching chunks of information at query time.

In this lab, you will connect to the Azure AI Search index from Python and run search queries programmatically. This shows how your application can retrieve relevant content from the index before sending that content to the model.

In this lab, we complete the full RAG loop. You will take a user question, search the index for relevant grounding data, pass both the question and the retrieved content to the model, and generate a more accurate answer based on your own documents.

In this lab, you will see how retrieval can be made available as a tool for an agent. This is an important pattern because it allows the agent to decide when it needs external knowledge and then use the search results to produce a grounded response.

This lecture introduces the pattern of building an agent that can call an external API using an OpenAPI tool. You will learn how this helps an agent move beyond conversation and connect to real business systems, such as order tracking, support platforms, or internal databases.

In this lecture, we explain the role of the OpenAPI specification. You will learn how the OpenAPI document acts as a contract that tells the agent what API operations are available, what inputs are required, and what responses can be expected.

This lecture introduces Azure Functions as a serverless way to host lightweight APIs and business logic. You will learn why Azure Functions are useful for agent scenarios where you need the agent to call a secure, scalable, HTTP-based backend service.

In this setup lecture, you will install Azure Functions Core Tools. This prepares your local development environment so you can create, run, test, and later deploy Azure Functions that can be used by your AI-powered applications and agents.

In this lab, you will build the CloudXeus Orders API using Azure Functions. This API will act as the backend system that the agent can call to retrieve order information, showing how agents can interact with real application data.

In this lab, you will publish the CloudXeus Orders API to Azure Functions. This moves the API from your local development environment into Azure, giving the agent a real HTTP endpoint that it can call when it needs to retrieve order information.

In this lecture, you will create the OpenAPI specification for the CloudXeus Orders API. The OpenAPI spec describes the available API operations, required inputs, response format, and security details so that Microsoft Foundry can understand how the tool should be used.

In this lecture, you will attach the CloudXeus Orders API to a Microsoft Foundry agent as an OpenAPI tool. You will see how the agent can use the API during a conversation to fetch real order data instead of guessing or relying only on model knowledge.

This lecture explains the authentication options available when using OpenAPI tools with agents. You will learn why authentication matters, how it protects backend APIs, and how the OpenAPI specification can define the security requirements for calling a tool.

In this lecture, you will secure the CloudXeus Orders API by using an Azure Function key. This shows a simple but important way to prevent anonymous access and ensure that only authorized callers, such as your configured agent tool, can invoke the API.

In this lecture, we pause to review the cost considerations for the solution we have built so far.

In this lecture, we introduce the concepts of state, conversations, session context, and memory. You will learn why these concepts matter when building agents that need to maintain continuity, understand previous messages, and provide a more natural user experience across interactions.

This lecture explains how memory works in an agent-based system. We look at the memory lifecycle, different types of memory, and the scope of memory so you can understand what an agent should remember, when it should remember it, and how that memory can be used safely and effectively.

In this chapter let's see how via code we can manage conversations for agents

This lecture explains how to give an agent access to memory by using memory stores and the Memory Search tool. You will see how stored information can be retrieved during a conversation so the agent can provide responses that are more personalized, contextual, and useful.

In this lecture, we introduce workflows in Microsoft Foundry. You will learn how workflows help coordinate multiple steps, connect agents together, apply conditions, and build more structured AI solutions instead of relying on a single conversational agent.

In this lab, you will create the CloudXeus agents that will be used in the workflow scenario. Each agent has a specific responsibility, helping you understand how multi-agent solutions can divide work across specialized agents.

In this lab, you will build a workflow in the Microsoft Foundry portal. You will connect agents together, define the workflow flow, and see how Foundry can orchestrate multiple agent actions as part of a single business process.

In this lab, you will capture the IntakeAgent’s output as structured data. This is important because structured outputs allow later workflow steps to make decisions based on fields such as request type, category, priority, or whether the customer needs escalation.

In this lab, you will add conditions to the workflow so the process can branch based on the IntakeAgent’s structured output. This shows how workflows can move beyond a fixed sequence and route requests dynamically depending on the user’s need

This lecture explains the operational side of running AI workloads. You will learn how quotas, scaling, rate limits, and cost affect real AI applications, and why developers must design solutions that are reliable, efficient, and cost-aware.

In this lecture, we take a quick look at where quotas and limits are managed. You will see how to review available capacity, understand service limits, and recognize why quota planning is important before building production AI applications.

This lecture introduces monitoring for AI workloads in Microsoft Foundry. You will learn how observability helps track agent behavior, application performance, failures, latency, and search health so issues can be detected and corrected early.

In this lecture, we introduce evaluators as tools for measuring AI quality. You will learn how evaluators help assess responses for aspects such as relevance, groundedness, completeness, safety, and usefulness in real-world AI applications.

In this lab, you will begin enabling continuous evaluation for an agent by connecting Application Insights. This provides the monitoring foundation needed to capture traces and collect the data required for evaluating agent behavior over time.

In this lab, you will create the continuous evaluation for the agent.

This lecture focuses on improving the completeness of AI responses.

In this lecture, we look at diagnostic settings and resource-level monitoring.

This lecture introduces agent tracing.

In this lecture, we introduce guardrails in Microsoft Foundry. You will learn how guardrails help reduce risk by controlling unsafe, harmful, or inappropriate outputs and why they are important when deploying AI agents in business environments.

In this lecture, you will see content filters in action using the support agent scenario. We look at how harm categories can help detect and manage unsafe content, making the agent more suitable for real customer-facing or internal support use cases.

This lecture introduces Prompt Shields and explains how they help protect AI applications from prompt injection and malicious instructions. You will learn why this matters when agents use external content, documents, tools, or user-provided inputs.

In this lecture, we introduce the Model Router in Microsoft Foundry. You will learn how model routing can help choose an appropriate model for a request, balancing factors such as capability, cost, latency, and task complexity.

In this lab, you will work with the Model Router in Microsoft Foundry. You will see how routing can be configured and tested so different requests can be handled by suitable models without hardcoding a single model choice for every scenario.

In this lecture, we introduce LangChain and explain why it is commonly used for building AI applications and agents. You will learn how LangChain helps developers connect models with prompts, tools, memory, and external data sources as part of a structured application workflow.

This lecture explains the core components that make up a LangChain agent. You will learn how the model, tools, prompts, agent logic, and execution flow work together so the agent can decide what action to take and how to respond to the user.

In this lab, you will set up the basic LangChain development environment and run your first simple example. This gives you a practical foundation before moving into full agent scenarios.

This lecture explains the process of building an agent with LangChain. You will see how an agent can receive a user request, reason about the task, choose a tool when needed, and generate a final response.

In this hands-on lab, you will build a working LangChain agent. You will connect the agent to a model, define tools, send test prompts, and observe how the agent decides when to use a tool to complete a task.

This lecture introduces LangGraph and explains why it is useful for building more controlled agent workflows.

In this lecture, we walk through the implementation design for a LangGraph-based agent.

In this lab, you will build an agent using LangGraph. You will define the graph, configure the agent flow, run the application, and see how LangGraph gives you more control over the steps an agent follows.

This lecture introduces the Model Context Protocol, or MCP. You will learn how MCP provides a standard way for AI applications and agents to connect to tools, services, and external systems without needing a custom integration pattern for every scenario.

In this lecture, we look at why MCP was created and what problem it solves. You will understand how MCP helps reduce integration complexity by giving agents a consistent way to discover and call tools across different platforms and services.

This lecture prepares you for working with an Azure MCP server.

In this lab, you will deploy an MCP server on Azure. This gives you a hosted endpoint that can expose tools to an AI application or agent, showing how MCP can be used in a cloud-based architecture.

In this lab, you will connect to the deployed Azure MCP server and test the available tools. You will see how a client can communicate with the MCP server and receive responses from the tools it exposes.

In this lab, you will build a custom MCP server. This shows how you can expose your own business logic, APIs, or backend operations as tools that agents can call during a conversation.

In this lab, you will test the custom MCP server you built. You will invoke its tools, verify the responses, and understand how a custom MCP server can become a reusable bridge between an AI agent and your own application logic.

In this lab, you will send an image as input to a chat model and ask the model to reason over the visual content. This demonstrates how multimodal models can work with both text and images, which is useful for scenarios such as support screenshots, product images, diagrams, and document-based questions.

This lecture introduces image generation with generative AI models. You will learn how text prompts can be used to create new images and why this capability is useful for scenarios such as marketing assets, product mockups, training visuals, and creative content generation.

In this lab, you will use Python to call an image generation model and create an image from a prompt. This gives you a developer-focused view of how image generation can be added to applications and automated workflows.

This lecture explains how generative models can be used not only to create new images, but also to modify existing ones.

In this lab, you will modify an existing image using a text prompt. The goal is to show how a model can apply a requested change while keeping the rest of the image as consistent as possible with the original scene.

In this lab, you will perform image editing using a mask. You will learn how a mask tells the model which area of the image can be changed, making this approach useful when you want precise control over what gets edited and what must remain unchanged.

This lecture introduces Prompt Shields for Documents and explains how they help defend against indirect prompt injection attacks

In this lecture, you will test Prompt Shields for Documents in a practical scenario.

In this lab, you will use guardrails for text moderation. You will see how text input or output can be checked for harmful content, helping developers build safer AI applications for customer support, internal assistants, and public-facing experiences.

This lecture introduces the idea of moderating image content.

In this lab, you will create an Azure AI Content Safety resource. This resource provides the foundation for checking text and image content for safety risks, helping you add responsible AI controls to applications and agent-based solutions.

In this lab, you will use Azure AI Content Safety to moderate image content.

In this lecture, we introduce how language models can be used for text analysis tasks. You will learn how models can extract meaning from text, identify important information, summarize content, classify messages, and support business scenarios such as support ticket analysis, feedback review, and document processing.

In this lab, you will use simple prompting to ask a model to identify entities and topics from text. This gives you a practical starting point for understanding how language models can perform text analysis before moving into more structured and specialized approaches.

In this lab, you will ask the model to return its response in a structured JSON format. This is important because AI applications often need predictable output that can be parsed by code and passed into downstream systems, workflows, or business applications.

In this lab, you will use a model to detect sentiment and tone from text. You will see how this can help organizations understand whether a customer message is positive, negative, urgent, frustrated, or neutral, which is useful for support prioritization and feedback analysis.

In this lab, you will use a model to translate text from one language to another. This demonstrates how language models can support multilingual communication and help applications serve users across different regions and languages.

This lecture introduces Azure AI Language and explains where it fits in the Azure AI ecosystem. You will learn how it provides prebuilt language capabilities such as entity recognition, language detection, sentiment analysis, key phrase extraction, and PII detection.

In this lab, you will use Azure AI Language to detect and redact personally identifiable information from text.

In this lab, you will detect the language of a given text using Azure AI Language.

In this lab, you will use prebuilt named entity recognition to identify entities such as people, locations, organizations, dates, quantities, and other important terms from text.

This lecture introduces the Azure Language MCP Server and explains how it can expose language capabilities as tools for AI agents.

In this lab, you will connect to the Azure MCP Language Server and use it from an AI application or agent scenario.

In this lab, you will list the tools exposed by the Azure Language MCP Server. This helps you understand how MCP clients discover available capabilities and how those tools can then be selected and used by an agent.

In this lab, you will use Azure Translator to translate text between languages.

This lecture introduces the speech and voice portion of the course. We set the context for working with speech-to-text, text-to-speech, real-time recognition, speech translation, voice-based agents, and the role of voice interfaces in modern AI applications.

This lecture explains speech-to-text and why it matters in AI applications. You will learn how spoken audio can be converted into text so it can be searched, analyzed, summarized, translated, or passed to an agent for further processing.

In this lab, you will use code to convert speech audio into text. This gives you a developer-focused understanding of how speech recognition can be added to applications such as call-center tools, meeting assistants, dictation apps, and support workflows.

In this lab, you will work with real-time speech recognition. You will see how audio can be captured and transcribed as the user speaks, which is useful for interactive voice applications, live captions, and conversational AI experiences.

This lecture introduces text-to-speech and explains how applications can convert written text into natural-sounding audio.

In this lab, you will generate spoken audio from text. You will see how to select a voice, send text to the speech service, and produce an audio file that can be used in applications or demos.

This lecture explains how speech can become a modality for agents. You will learn how the Azure Speech MCP Server can help connect speech capabilities to an agent so users can interact through voice rather than only through typed text.

In this lab, you will connect the CloudXeus agent to the Speech MCP Server. This shows how an agent can use speech capabilities as tools, helping create a more natural voice-enabled support or assistant experience.

This lecture introduces the Voice Live API at a high level. You will learn how real-time voice interactions can be used to build conversational experiences where users speak naturally and receive responses with lower latency.

In this demo, you will see the Voice Live API in action. The goal is to understand how live voice input and output can support more interactive AI experiences compared to traditional request-and-response text-based applications.

This lecture compares speech translation with LLM-based speech experiences. You will learn when to use a dedicated speech translation service and when a language model-based approach may be more suitable for reasoning, summarization, or conversational experiences.

In this lab, you will translate spoken audio from one language into another. This demonstrates how speech translation can support multilingual communication in scenarios such as meetings, support calls, training, travel, and global customer interactions.

In this lecture, we introduce Azure Content Understanding and explain where it fits in the Azure AI ecosystem. You will learn how it can analyze different types of content, extract meaningful information, and support scenarios such as document processing, image understanding, and knowledge extraction.

This lecture explains the key document analysis capabilities used in Azure Content Understanding. You will learn how OCR extracts text, how layout analysis identifies structure, how tables are detected, and how fields can be extracted from documents such as invoices, forms, and reports.

In this lab, you will use Azure Content Understanding to analyze a document. You will see how the service reads the document, extracts text and structure, and returns useful information that can be used by applications, workflows, or agents.

In this lecture, we review the document analysis workflow and reinforce the important concepts.

In this lab, you will use Python to analyze an invoice with Azure Content Understanding. You will see how code can submit the invoice, retrieve the analysis result, and work with extracted fields such as invoice number, vendor details, dates, totals, and line items.

This lecture introduces visual understanding with image analyzers. You will learn how Azure Content Understanding can analyze images and extract useful information from screenshots, product images, diagrams, forms, and other visual content.

In this lab, you will use an image analyzer to process visual content.

This lecture explains the difference between Standard mode and Pro mode pipelines in Azure Content Understanding. You will learn how different pipeline modes affect how content is processed and why selecting the right mode matters for accuracy, complexity, and business requirements.

In this lecture, we connect Azure Content Understanding with agent-based scenarios. You will learn how extracted information from documents or images can be passed to an agent so it can summarize, reason, answer questions, or take the next step in a workflow.

In this lab, you will use Content Understanding together with an agent. You will analyze content, pass the extracted result to the agent, and generate a useful response based on the information found in the document or image.

This lecture introduces Azure AI Search as a key service for Retrieval Augmented Generation. You will learn how search helps retrieve relevant business content so a model or agent can produce grounded answers using your own documents.

This lecture explains the core building blocks of Azure AI Search. You will learn what a data source is, how an index stores searchable content, and how an indexer moves data from a source into the search index.

In this lab, you will prepare documents stored in Azure Blob Storage for ingestion into Azure AI Search. This demonstrates a common enterprise pattern where files are stored in a container and later indexed for search or RAG scenarios.

In this lab, you will create or configure a data source in Azure AI Search. This data source tells Azure AI Search where your content is located and prepares the foundation for automated indexing.

In this lab, you will create the Azure AI Search index. You will define the fields that store your searchable content and understand how the index structure affects search, filtering, retrieval, and later RAG integration.

In this lab, you will create the indexer that connects the data source to the index. The indexer is responsible for reading the source content, processing it, and populating the search index with searchable data.

In this lab, you will run queries against the Azure AI Search index.

This lecture introduces AI enrichment and skillsets in Azure AI Search. You will learn how skillsets can add intelligence during indexing, such as OCR, key phrase extraction, language detection, and other transformations that make content more useful for search and RAG.

In this lab, you will configure AI enrichment by using skillsets. You will see how documents can be enhanced during indexing so the final search index contains richer, more searchable, and more useful information.

This lecture explains how Azure AI Search can be connected to an agent as a knowledge source or retrieval tool. You will learn how this allows the agent to retrieve relevant content from indexed documents and generate responses grounded in enterprise data.

In this lab, you will connect Azure AI Search to an agent and test the full experience. The agent will use the search index to retrieve relevant information and respond based on the indexed content instead of relying only on the model’s built-in knowledge.

This chapter is about the Introduction to this new course on AI 102

This chapter is a brief understanding on Machine Learning

This chapter briefs on Generative AI

This chapter talks about the Hot topic :Chat GPT

This chapter is about  Large Language Models

This chapter is about  GPT(Generative Pre-Trained Transformer)

This chapter explains the model lineup in OpenAI.

This chapter is about Multi-modal LLM

This chapter shows how to sign-up to the OpenAI platform.

This chapter shows how to use the playground at the OpenAI platform.

This chapter shows how to use system messages at the OpenAI platform.

This chapter talk about the next steps

This chapter is a brief introduction to Azure

This chapter briefs about Azure Free account

This chapter briefs on creating the Azure free account

This chapter gives a tour around the Azure portal

This chapter is about Azure AI Services

This chapter is on AI Safety and Responsibility

This chapter gives a brief on Azure OpenAI and Azure AI Foundry

This chapter provides overview on Deploying an Azure AI Foundry Resource.

This chapter demonstrates how to creating an Azure AI Foundry Resource

This chapter briefs about Launching the Azure AI Foundry portal

This chapter briefs on exploring the models in Azure AI Foundry

This chapter gives a brief on how to deploy a model in Azure AI Foundry

This chapter briefs on a Lab on deploying a model via Azure AI Foundry.

This chapter explains how to use the playground in Azure AI Foundry.

This chapter briefs on using model parameters in Azure AI Foundry.

This chapter briefs on using the system prompt in Azure AI Foundry.

This chapter describes how to generate images in Azure AI Foundry.

This chapter briefs about the Prompt Engineering - Basic Prompt fundamentals

This chapter briefs about the Prompt Engineering - Zero-shot Prompting

This chapter explains on Prompt Engineering - Few-shot Prompting

This chapter draws the attention on the topic , Prompt Engineering - Chain-of-Thought Prompting

This chapter is about Installing Python

This chapter is about Installing Visual Studio Code

This chapter shows how to run a simple Python program

This chapter shows how to use the Chat Completion API from within the Python program.

This chapter shows how to use system messages from the Python program.

This chapter shows how to generate the entire response from the Python program.

This is a Quick note on monitoring the model usage

This chapter showcases the multi-modal support available within models.

This chapter shows how to ask the model to explain code

This chapter shows how to use a model to generate an image from within Python.

This chapter is about Fine Tune Azure Open AI models

This chapter provides overview of Fine tuning a model

This chapter briefs about Retrieval Augmented Generation

This chapter shows the step of using the Azure Blob service in Retrieval Augmented Generation.

This chapter shows the step of using the Azure Search service in Retrieval Augmented Generation.

This chapter is about the complete implementation around Retrieval Augmented Generation.

This chapter is about using Retrieval Augmented Generation from Python.

This chapter is about Azure AI Content Safety

This chapter shows how to add a content filter in Azure AI Content Safety.

This chapter is a Quick note on Azure AI Content Safety - BlockList

This chapter shows how to create an Azure AI Content Safety resource.

This chapter shows how to use the Azure Safety Content Studio.

This chapter shows how to use the Prompt Shield in Azure AI Content Safety Studio.

This chapter shows how a Python program can be used to detect harmful images using the Azure AI Content Safety as a service.

This chapter shows how a Python program can be used to detect harmful text using the Azure AI Content Safety as a service.

This chapter shows how to make REST API calls against the Azure AI Content Safety as a service.

This chapter gives an introduction to the Computer Vision service.

This chapter shows how to create a computer vision resource

This chapter shows to perform image tagging in Computer Vision.

This chapter shows to perform image tagging in Computer Vision and Python.

This chapter shows how to generate captions in Computer Vision and Python.

This chapter shows how to perform object detection in Computer Vision and Python.

This chapter shows how to perform optical character recognition in Computer Vision and Python.

This chapter shows how to perform brand detection in Computer Vision and Python.

This chapter is about the using VS Code to make API calls.

This chapter briefs on The Face service

This chapters shows in brief how to use the Face service.

This chapter shows how to use the Face service from Python.

This chapter is on Azure AI Custom Vision

This chapter shows how to create the Azure AI Custom Vision resources.

This chapter shows how to perform image classification in Azure AI Custom Vision.

This chapter is a Quick understanding on the training results

This chapter shows how to perform image classification in Azure AI Custom Vision and Python.

This chapter shows how to perform object detection in Azure AI Custom Vision.

This chapter is on Azure AI Video Indexer

This chapter shows in brief how to use the Azure AI Video Indexer service.