Create custom shipping rates using AI tools - Build the feature

Transcript

Hi, this is Russell with Adobe. This is a recorded demo from an Adobe engineer on how AI-powered agents can be used throughout the development workflow. We’re going to pick up with our speaker and where he’s explaining role-based reasoning and how this starts to shape the architecture and implementation plan. We first run an AIO login. This is required so we can connect through the MCP tool that provides access to the RAG tooling. I already performed this step earlier and now we’re ready to start Clod.

The agent initializes with an AgentMD file, which gives it context about the task it’s about to perform. In this case, the task is to act as a specialist in checkout starter kit integrations.

We also review the available MCP tools, the skills, and the files it has in memory. Focusing on the MCP tools, these are the different commands we can access. So that is basically what we package with our tooling. Now let’s talk about the use case we’re going to build.

We’re going to extend the commerce checkout with external shipping rates coming from an outside system. To avoid tying the solution to any specific commercial provider, we’re using a mock API that simulates this type of service. The API is secured with an API key and the input it expects is the set of items that would be part of a checkout cart. It responds with the corresponding shipping rates. We’ll open the boilerplate connected to the commerce system where we will make the changes.

You’ll see that currently only the standard shipping rate is available when you have a clean ACCS instance.

This is our starting point. From here, we’re going to ask the agent to create an extension that adds support for the new shipping methods provided by the external system. To give the agent proper context, we also documented the API it needs to use in the mock rating API spec, which we added to the workspace. We could have described this in natural language, but providing the actual specification is It includes the payload details, HTTP method, and other necessary information.

At this point, we’re ready to start integrating and extending the checkout with these externally provided shipping rates. This is the prompt we are using for the example. It essentially says, let’s build another ecommerce extension that adds metadata to checkout. The shipping rates come from the external shipping rate service, and we want to make the service URL and API key configurable in admin.

So we’re adding this configurability requirement, and we expect the agent to generate an admin UI SDK component to extend the commerce admin UI. We also provide details about the shipping API, such as the service endpoint and the API key, both of which should be configurable. We also provide security guidance. We don’t want secrets to appear in any logs. With that, the agent begins reasoning, and the expectation is that it will use the available skills to go through the different phases of the software development cycle.

The first phase is the product manager role. It gathers requirements, understands the scope of the project, and documents the requirements file. To do that, it’s instructed to ask clarifying questions whenever something is unclear.

These are the kinds of questions it asks.

What is the target environment, PaaS or SaaS? We select both.

Should the extension return shipping rates conditionally or always? We choose conditional, only when applicable.

It also asks about caching, whether we want to cache rates or not. We choose not to use cache for now. It asks about generating tests. In this example, we skip test generation and submit the response.

With these questions answered, the agent generates the requirements document, which becomes the source of truth for the solution it will implement. It then moves on to the next phase, which is architecture.

To do that, we’ll set up the session for creating custom shipping rates with Adobe’s AI tools. There are a lot more videos on this topic that you can find on Experience League, and I hope you continue to come back to Experience League to learn more about Adobe Commerce, as well as all of the other Adobe products.