Building an AI Agent Platform in .NET 10 with GitHub Copilot

2:00 下午 - 3:00 下午 (UTC)

In this session we will transform our chatbot into an intelligent agent that can call tools to interact with the outside world. We will build a tool framework with ITool, IToolRegistry, and IToolExecutor interfaces, and implement production-ready tools for file system access, shell commands, web fetching, and task scheduling — all with built-in security guardrails. At the end we will have an agent that reasons through multi-step tasks on its own. Ask it to "Summarize my project's README" or "Check what's in my Downloads folder" — it will break the problem into steps, call the right tools, and report back. Key Topics & Technologies: Agent orchestration and reasoning loops Tool abstraction and design patterns Security considerations (path traversal, command injection, SSRF) Tool UI and result streaming LLM function calling protocol Dependency injection and service registration Structured logging and observability hooks

2:00 下午 - 3:00 下午 (UTC)

In this session we will personalize the agent with skills — reusable knowledge bundles defined in Markdown with YAML frontmatter — and add conversation memory so it retains context across sessions. We will build a skill loader, implement automatic summarization to manage token budgets, and create a UI to toggle skills on and off in real time. At the end we will have an agent that adapts its expertise based on loaded skills and remembers your preferences across conversations. Add a "Git Expert" skill or a "JavaScript Mentor" skill and watch the agent's behavior change. Your users can create custom skills without touching code. Key Topics & Technologies: Skill architecture (Markdown + YAML) Markdown parsing and metadata extraction In-memory caching and skill composition Conversation memory and summarization strategies Token budgeting and context window management Skill UI with toggle controls Workspace customization patterns

2:00 下午 - 3:00 下午 (UTC)

In this session we will make the agent production-ready. We will add job scheduling as a background service, integrate Azure OpenAI and Azure AI Foundry as cloud model providers, build monitoring dashboards, and write unit tests that verify the entire stack. At the end we will have a complete, tested, cloud-ready agent platform with production-grade error handling, multiple AI model providers, scheduled automation, and observability dashboards. Deploy it to Azure Container Apps, App Service, or run it on-premises with Ollama. Key Topics & Technologies: BackgroundService for long-running jobs Azure OpenAI integration and model selection Azure AI Foundry (inference, deployments, evaluation) Unit testing patterns for agents and tools Health checks and system diagnostics Job queuing and status tracking Configuration for multi-environment deployments (local, staging, production)