Live Coding Second Breakfast: Building Smarter AI Agents for the Garden Now Project (Part 2)

[Kristy Merriam]

This week on Live Coding Second Breakfast, we jumped back into our AI agent for the Garden Now project - a community-focused smart garden tool built with ServiceNow. With Laszlo Balla joining the stream, we explored how to use MetricBase, AI Agent Studio, and a bit of scripting to make plant care smarter and more automated.

After some banter about coffee and time zones, Kristy and Laszlo tackled the next problem: why wasn’t the AI agent showing up in the Agent Studio interface, even though it worked during testing? Turns out, the agent was a bit camera-shy (or maybe just bugged).

Once that was sorted, the team moved on to enriching the agent’s capabilities by allowing it to access up-to-date sensor readings from MetricBase. With ServiceNow’s Workflow Studio and a utility Script Include, they created a new tool that could fetch the latest soil moisture, air temperature, acidity, light, and more from a metric-based data source.

This led to a lively discussion on how MetricBase works, when to use it (like with Smart Kegs!), and how to manage sensor data over time. They walked through writing a reusable utility function, building a dynamic script action that returns sensor readings as an array of objects, and dealing with the ever-classic [object Object] problem in output mappings.

The episode wrapped up with a win: sensor data retrieved, debugged, and ready for the AI agent to use in future decisions.

This is what real dev work looks like: figuring things out as you go, asking "Should this be an array?", and making progress one step at a time.

Join us next week where we'll pick up where we left off an hopefully be able to test the AI Agent using these new actions.