Microsoft Semantic Kernel: from Zero to Hero скачать в хорошем качестве

Microsoft Semantic Kernel: from Zero to Hero

Recording of presentation at the BC TechDays 2025 conference of 12 & 13 June 2025 in Antwerp (Belgium) Title: Microsoft Semantic Kernel: from Zero to Hero Presenter: Stefano Demiliani Presented on Thursday 12 June 2025, 16:00 - 17:30 TOC: 00:00 - Introduction 00:42 - Session Goals and AI Solution Requirements 02:57 - Challenges in AI Adoption by Businesses 05:16 - Characteristics of Enterprise-Ready AI 06:55 - Overview of AI Frameworks and Microsoft’s Semantic Kernel 08:13 - LLM Basics: Tokens, Prompts, Memory, and Reasoning 10:14 - Memory Types: Short-Term vs Long-Term (Vector DBs) 12:26 - Agents and Tool Use in the Microsoft AI Ecosystem 14:05 - Limitations of Copilot Studio and Why Full Code Matters 15:04 - What is Semantic Kernel and Why It’s Useful 16:59 - Semantic Kernel Components and Supported Languages 18:38 - Connecting to Multiple LLM Providers 21:00 - Basic Chat Completion Example (Azure OpenAI) 23:17 - Maintaining Conversation Context Across Prompts 25:01 - Switching Between AI Providers (e.g. Google Gemini) 27:16 - Introduction to Vector Databases and RAG (Retrieval-Augmented Generation) 29:36 - Using Kernel Memory for Document Chatting 32:05 - Setting Up Memory with File System, Azure SQL, and More 34:58 - Live Demo: Ask Questions Against Business Central Documentation 37:43 - Implementing Semantic Kernel Plugins (Prompt Templates & Native Functions) 40:50 - Using Azure Logic Apps as AI Plugins 44:00 - Demo: Native Functions as Tools 48:58 - Demo: Invoking Logic Apps via Swagger from Semantic Kernel 50:37 - Live Workflow Orchestration Using Azure Logic Apps 54:57 - Agent Patterns Overview (Sequential, Concurrent, Group Chat, End-of) 57:20 - Implementing Sequential Agent Orchestration (Product Description Pipeline) 1:00:58 - Concurrent Agent Pattern (Company Info + Stock Market Example) 1:06:02 - End-of Pattern: Order Fulfillment Use Case with Business Central 1:11:18 - Human-in-the-Loop Interaction in Agent Flows 1:13:03 - Using Semantic Kernel with Business Central via Azure Functions 1:16:00 - Complex Queries and Analysis in Business Central via AI Agents 1:18:43 - Dynamic Prompt Generation Based on Business Central Metadata 1:21:02 - Teaching the Model to Use APIs Through Prompt Instructions 1:23:14 - Model Contract Protocol (MCP) Overview and Usage 1:25:53 - Implementing MCP Support in Semantic Kernel 1:26:57 - Playwright MCP Demo: Browsing and Summarizing Web Pages 1:28:43 - Future of Microsoft Agent Connectors and MCP Adoption 1:30:31 - Summary: Pros, Cons, and Use Cases for Semantic Kernel in Business Central 1:32:10 - Q&A