Super Agent to create AI Agents

Welcome.

Let's dive into our incredible feature the Super Agent, which can create AI agents.

Here's our summary dashboard displaying the total agent creation plans made, what's scheduled in development and completed.

This dashboard offers a detailed look at each plan.

Let's go ahead and craft a new one using my initial prompt.

Meanwhile, we can explore existing plans.

Let's check out one called AIOps alerts agent, our main landing page here.

boys.

Here we provide an in depth analysis.

This first overview page reveals the number of tools created, types of tools, and total test cases executed for this agent.

You can view a flow diagram indicating timestamps and which agent was executed.

Our first phase focuses on the Planner agent.

among the available tools we have the Query Stream tool and for it we need the getalerts and Get Alerts Metadata tools.

Importantly, no credentials are needed.

Our Planner agent generated a detailed plan showing no available agents, then checked available tools.

There is only one available tool.

The Required Tools section details the toolset name, description, expected input parameters, output, format fields provided, and main tool type.

LLM offers this comprehensive plan, highlighting the second tool GetAlerts data and its parameters.

It confirms no credentials are needed, identifying Query Stream as an available tool.

It also outlines this existing tool's basic test cases for execution verification, including positive cases, edge cases, and error scenarios handling.

Similarly, detailed plans are provided for all three required tool types.

Moving forward, this is our comprehensive execution plan.

It indicates the need to create two new tools alongside an existing one.

After completing steps 1, 2 and 3, we proceed to create and use the agent, considering key integration points.

This step by step guide for the entire agent details how the plan was submitted to the tool creation agent, resulting in the creation of two required GetAlerts data and get Alerts metadata.

Click View Details to see the tool definitions and descriptions.

You can confirm result columns, input parameters like stream name, severity, and status, all recorded by LLM.

Next, execute test cases on the newly created and existing tools.

During tool verification, review the comprehensive list of executed test cases.

Analyze details about invalid syntax, current states, and whether tests passed, failed or represented, negative, positive or edge cases.

Results are available and you can click on View Details.

Let's examine a failed test case that later passed for clarity.

Filter by tool name to access data.

Once tools are created and verified, the next step is agent creation.

In agent creation, a new Persona is crafted incorporating new tools, tool sets, and prompt templates.

Click Cross Launch.

The Persona is pre filled, selecting the appropriate LLM.

Click Start to see sample prompts that you can furnish to use the agent showcasing the agent's capabilities.

We will select one of the prompts and the agent is ready to go.

This concludes the demonstration.

Thanks for joining us.

Catch you in the next video.