Welcome to the Semantic Kernel Agent Lab!
This lab's objective is to provide an introduction on how to build an agent using Semantic Kernel, add specific skills to the agent for executing actions to solve problems, and finally, create a team of agents to collaboratively solve more complex problems.
This lab contains three exercises to help you gain hands-on experience with Semantic Kernel Agents.
At the time of this workshop, The Semantic Kernel Agent Framework is experimental, still in development and is subject to change. All documentation is Semantic Kernel Agent Framework (experimental).
The objective of the first lab exercise is to create an agent with reasoning capabilities to solve domain-specific requests. To achieve this, the agent utilizes a Large Language Model (LLM).
For detailed instructions, please refer to the Basic agent with SK.
In this exercise, you will create your first Semantic Kernel agent that has a skill to get the current weather of a city by calling a public API.
In the context of Semantic Kernel, agent skills refer to the capabilities or functions that an AI agent can perform to solve specific tasks. These skills enable the agent to interact with its environment, process information, and provide meaningful outputs.
In Semantic Kernel, an Agent Group Chat is a framework that enables multiple AI agents to interact and collaborate within the same conversation. This setup allows agents to work together, leveraging their individual skills and capabilities to achieve a common goal1.
In this exercise, you will explore build a team of 4 agent that going to work togther to solve an user trip request.
Happy coding!
- All the exercises are built in Visual Studio Code but you could use Visual Studio as well
- AzureOpenAI LLM Endpoint and Key
- Last .NET SDK installed
Building agents with Semantic Kernel.
- Create an agent with reasoning capabilities to solve domain-specific requests
- Add specific skills to the agent for executing actions to solve problems
- Create a team of agents to collaboratively solve more complex problems
- Experiment with different agent skills and their interactions
- Explore the impact of different agent collaboration strategies