From the course: Creating Agents with Python and the AI Toolkit for Visual Studio Code

Overview of the AI Toolkit panels and commands

From the course: Creating Agents with Python and the AI Toolkit for Visual Studio Code

Start my 1-month free trial Buy for my team

Overview of the AI Toolkit panels and commands

“

With the AI Toolkit for Visual Studio Code extension installed, let's now explore the AI Toolkit. On the left-hand side in the activity bar, you'll select the AI Toolkit icon. You may need to go to additional views, and then you can select AI Toolkit. That'll bring you directly to the extension. I'm just going to close the extension page from there. Okay, starting with the left-hand side here where we have the toolkits panels. So at the very top, we have my resources, and that's going to be where you can find any models that you've deployed, any agents you've created, as well as any MTP servers that you have access to, or that you've added to the toolkit, I should say. Then we have our model tools, and our model tools is where you're going to find model catalog, the model playground. If you have done any model conversion, you'll find the instructions there or the area where you can do that model conversion, and then also we have some fine-tuning capabilities in here as well. Just below, we have the agent and workflow tools, and that's where you're going to find the agent builder, which is where you can go and create your agent. You can also find the bulk running of evaluations here as well, and also our tracing features here for our tracing viewer. Then down below, we have a section for building an agent with GitHub Copilot. You can use GitHub Copilot to help you create your agent for a code-first version of creating your agent. Then just below within the MCP workflow, this is where you're going to find your area to either add an MCP server or create a new MCP server. Many of these areas we'll explore throughout this course, but I'll take you on a tour of what some of these look like now. Within the model catalog, this is where you can come and find models to learn more about their capabilities, as well as where you can go to deploy them, or you can also try those directly in what is called the model playground. The model playground is where you can come and chat with your chosen models. Within here, you can submit your prompts, and you can also modify any model preferences as well. We then have the Agent Builder. The Agent Builder is where we're going to be working today to create ourselves inside Agent. I happen to have one up here already. But within the Agent Builder itself, this is where you can come and configure your agent in this UI workflow. You can also chat with the agent in the playground, and then you can also run some evaluations on your agent as well, and we'll explore that later. There's also the tracing feature, which is where you can come and view the traces for your agents as well. You can find information here such as the start time, the duration, as well as the total tokens that were used as well. All of these panels work together. There's the model catalog, as I mentioned earlier, and that's for finding models. There's the playground, which is for testing and comparing models. The agent builder for creating an agent and assessing the agent output. Then finally, there's the evaluations, which are part of the agent builder. Now finally, tracing. This is where you can come and do debugging for your agents. You

Contents