A starter kit for creating AWS Lambda microservices with GitHub Copilot for AI assisted pair programming.
Begin by reviewing and updating the Copilot Instructions to suit the needs and preferences for your project.
Interact with Copilot using your favorite editor to create your application. We strongly recommend VS Code with the following extensions:
- GitHub Copilot
- GitHub Copilot Chat
- Prettier - Code formatter
- ESLint
- Vitest
- indent-rainbow
The official guide for GitHub Copilot in VS Code provides a lot of information and is very useful for those who are just starting to use Copilot in VS Code.
The VS Code YouTube channel has a playlist with many interesting videos and tutorials for GitHub Copilot in VS Code.
This official tips and tricks guide from VS Code provides an excellent summary of Copilot best practices.
The project structure follows the best practices for Copilot assistance. This project is ready to begin iterative development with a Copilot AI agent.
Note that the application component does not exist, not yet anyway. This base project structure is primed for building an AWS Lambda component from the very beginning using Copilot agent mode with you in the driver's seat.
/.github
copilot-instructions.md # Copilot instructions
/docs
/requirements
01-create-task.md # Requirements specification
.editorconfig # Editor configuration
.prettierrc # Prettier configuration
.gitignore # Git ignore
LICENSE # Software license
README.md # This document
To get started, you should fork this repository on GitHub or copy it to your local machine. This allows you to make changes, track your progress, and customize the project to fit your needs. If you are new to GitHub, you can use the 'Fork' button at the top right of the repository page to create your own copy.
📢 Please do not push commits to this repository. This repository is for teaching purposes only.
Add a section to the Copilot Instructions document immediately following the Role section. Provide an overview of the project like this:
---
## Project Overview
- **Component:** Task Service **task-service**
- **Description:** This service provides a REST API for managing tasks, including creating, retrieving, updating, and deleting tasks. It uses AWS Lambda functions triggered by API Gateway events, with business logic encapsulated in service classes. The project follows best practices for TypeScript development, AWS CDK infrastructure management, and unit testing with Vitest.
---Want to learn more about Copilot instructions files? Read more in the official guide...
Begin working with Copilot to create the application. Copilot works best when it has a rich context to use as reference material when updating the project. When you first begin, there is no source code for Copilot to reference. Working with Copilot to implement the first story may seem tedious for this reason. However, I find it best to let Copilot implement the first story to the best of the agent's ability. Then, review all of the generated source members. Make changes to the source to match your coding preference. Don't like to use class? Change it. Prefer default exports? Use those.
When you implement the second and subsequent stories, Copilot will use the existing code as a reference and pattern the code that the agent creates to match.
💡 The "Claude Sonnet 3.7" model has been used to develop this project. Language models evolve quickly and new versions published frequently. Review the official documentation and select the language model that is most appropriate for your application or task.
The requirements are located in the docs/requirements directory. Open Copilot chat in VS Code and place it into Ask mode if it is not already.
Drag the 01-create-task.md requirements specification into the Copilot chat. Any files you explicitly drag onto the Copilot chat will be included in the context of the current session. Enter the following prompt:
let's update the project with these requirements.
OR
let's create a step-by-step plan to implement these requirements in the project.
or you can use the #file command to explicitly reference the requirements like this:
let's update the project with the requirements in #file:01-create-task.md
Since Copilot is in Ask mode, it will not make any changes to the project. Instead it will tell you what the agent would do if it were in Agent mode. Ask mode allows you to ask Copilot questions or, in this instance, allows you to get a preview of a more complex set of changes.
If you do not like the approach that Copilot proposes, update the requirements specification to be more specific. Remember that Copilot will improve as the code base grows.
Change Copilot from Ask to Agent mode. In Agent mode, Copilot is free to make changes to your project. Don't worry though, you may pause or cancel the Copilot session at any time. Any changes made by Copilot are not fully applied until you choose to Keep them, or you may revert the project by choosing to Undo the changes.
When you are ready, submit the prompt in Agent mode and the Copilot agent will update the VS Code project with the changes.
Copilot often works best when asked to implement small units of work. For that reason, work on 1 story at a time. When Copilot has finished, you play the role of the code reviewer (you are the expert after all). Review and update the code as needed (or ask Copilot to make a change). Run the unit tests. Run the linter. Deploy and test the changes. Finally, create a PR and commit the work.
Then, move to the next story and repeat.