Azure AI Agent Service - In 15 Mins

Talk given at the AI Dev Tools meetup in NYC on Feb 18
https://lu.ma/7wn1tmc0

Abstract:
A rapid introduction to the Azure AI Agent Service - what it is, why it matters, and how to create your first agent with built-in action and knowledge tools for a real-world application scenario.

Key Takeaways
Azure AI Agent Service is a MANAGED service in Azure AI Foundry
Azure AI Agents are SMART MICROSERVICES that can retrieve knowledge and take action
ACTION TOOLS like Function Calling and Code Interpreter streamline task automation
KNOWLEDGE TOOLS like File Search and Bing Search help ground responses in relevant data

Lab:
https://aka.ms/aitour/WRK552

Transcript

1. A I A G E N T D E V T O O L S - 0 2 / 2 5 B U I L D I N G A I A P P S W I T H T H E A Z U R E A I A G E N T S E R V I C E Nitya Narasimhan, PhD Senior AI Advocate https://linkedin.com/in/nityan
2. 1. What is an AI Agent? 2. Why use the Azure AI Agent Service? 3. Scenario: Contoso Sales Assistant 4. Setup: Create My First Agent 5. Action: Function Calling (SQL query) 6. Action: Code Interpreter (Python) 7. Knowledge: Document (File Search) 8. Knowledge: Real-Time (Bing Search)
3. What is an AI Agent? What Can I Use Them For? An AI Agent is semi-autonomous software designed to achieve a given goal without requiring predefined steps or processes. Instead, it determines what to do dynamically. In the Azure AI Foundry platform, an AI Agent acts as a "smart" microservice that can be used to answer questions (RAG), perform actions, or completely automate workflows. Show: Foundry Portal https://aka.ms/aitour/ai-agents-theater-session
4. What is the Azure AI Agent Service? Why use the Azure AI Agent Service? It is a fully managed service for developers to securely build, deploy, and scale high-quality, and extensible AI agents without having to worry about compute & storage. Streamlined and secure way to build AI agents with automatic tool calling (server-side), securely managed data (conversation state) and rich built-in tools for data, actions, RAG. Show: Documentation https://aka.ms/aitour/ai-agents-theater-session
5. Scenario: Contoso Sales Assistant Development: Code-First or Low Code A conversational agent designed to answer questions about sales data, generate charts for understanding, and download data files for reporting and analysis – for an enterprise. Create tailored AI agents by using custom instructions, custom functions, and tool-driven augmentations – using either the UI (portal) or the SDK (Python, C#, JS/TS) Show: AI Tour Workshop https://aka.ms/aitour/ai-agents-theater-session
6. Provision: An Azure AI Foundry Project Create: My Contoso Sales Assistant Configure project client with connection string. Then use it to create agent with name, model, tools and instructions and specify any tool-related resources required. Show: Azure AI Project A standard agent using Managed Identity for authentication – with key vault, storage, and AI Services (e.g., AI model deployment) along with AI hub and project resources. https://aka.ms/aitour/ai-agents-theater-session
7. Action Tool: Function Calling Use Case: Query My SQLite Database Function calling enables LLMs to interact with external systems, execute tasks, and integrate with APIs. Define a function, add it Function Tool toolset. It gets invoked for you. Enable function logic to execute dynamic SQL queries against a SQLite database in 3 steps. Define function, add function to agent toolset, update agent instructions to trigger tool use. Demo 1: Answer Query https://aka.ms/aitour/ai-agents-theater-session
8. Action Tool: Code Interpreter (Python) Use Case: Visualize result as a chart Code interpreter allows agents to write and run Python code in a sandboxed execution environment. Agents can be configured to try different iterations (on failure) till it succeeds. ”Show sales by region as a pie chart” – first runs basic function call to get SQL results, then writes Python code to visualize data and store it in a local file for user viewing. Demo 2: Create Charts https://aka.ms/aitour/ai-agents-theater-session
9. Knowledge Tool: File Search Use Case: Grounding with Documents ”Chat with your data” – the agent can answer questions on information found in uploaded files. It uses built-in vector store capability to parse and index information for later retrieval. Uses the File Search tool to retrieve relevant information to enable RAG-style search on a PDF file with product data. Tool uses a vector store with built-in semantic search capability. Demo 3: Use PDF Data https://aka.ms/aitour/ai-agents-theater-session
10. Knowledge Tool: … Bing Search Use Case: Grounding in Real Time data The Grounding with Bing Search tool allows you to incorporate real-time public web data in returned responses. Create the resource & connect it. Agent decides if & when to use it Provide instructions to agent to use tool for competitive insights and present concisely. We can’t access raw content but use and display requirements include query link & citations. Demo 4: Search Results https://aka.ms/aitour/ai-agents-theater-session
11. B U I L D I N G A I A P P S W I T H T H E A Z U R E A I A G E N T S E R V I C E STAR THIS REPO Found this useful? Star the repo and check back for the slides, visual and more learning resources!
12. Open-Source Curriculum Self-Guided Workshop https://aka.ms/aitour/wrk552 https://aka.ms/ai-agents-beginners Self-Guided Learning