AWS re:Invent 2025 - Supercharge Serverless testing: Accelerate development with Kiro (CNS427)

Supercharging Serverless Testing with Kiro: Accelerating Development at AWS re:Invent 2025

Challenges with Serverless Applications

• Serverless applications are highly distributed and modular, with a larger number of integrations compared to traditional apps
• This makes it important to test the integration layers, and raises questions about how to handle dependencies for isolated tests
• Lambda functions may not be complex enough, making it difficult to determine where to focus testing efforts

Evolving the Codebase for Testability

• The initial codebase was tightly coupled to infrastructure choices, making it difficult to change the underlying AWS services without rewriting tests
• The developer experience required detailed knowledge of mocking the AWS service APIs
• There was room to improve the speed and efficiency of the tests

Leveraging Dependency Injection and In-Memory Fakes

• Introduced dependency injection to decouple the handler code from the domain logic and integration layers
• Created configurable in-memory fake services to simulate different test scenarios, rather than using brittle mocks
• This simplified the test code and made it more resilient to changes in the domain logic

Property-Based Testing for Algorithmic Correctness

• Used property-based testing with the Hypothesis library to validate the algorithmic correctness of the circular dependency check
• Generated a large number of random test cases to thoroughly exercise the logic, rather than relying on a few fixed inputs
• Provided detailed visualization and debugging capabilities to understand the generated test cases and identify issues

Integration Testing Against Real AWS Services

• Tested the integration with DynamoDB and EventBridge using the actual AWS services, rather than mocks
• Validated not only the happy path scenarios, but also the error handling and exception surfacing
• Ensured the application behaved correctly when encountering integration failures, such as conflicts during offline task updates

Automating the Development Lifecycle with Kiro

• Leveraged Kiro's custom agents and spec-driven development approach to automate the review, planning, and execution of the architectural changes
• Used Kiro's property-based testing capabilities to generate comprehensive test suites for the domain logic
• Integrated Kiro's bug analysis and risk heatmapping features to continuously monitor and improve the application's quality in production

Key Takeaways

• Decouple the application architecture to simplify testing, using techniques like dependency injection and in-memory fakes
• Leverage property-based testing to validate the algorithmic correctness of critical business logic
• Test against real AWS services to ensure the application handles integration failures gracefully
• Automate the development lifecycle with tools like Kiro to streamline the review, planning, and execution of architectural changes
• Continuously monitor and improve the application's quality by analyzing production data and risks

Technical Details

• Programming language: Python
• Testing frameworks: Pytest, Hypothesis
• Mocking library: Moto
• Dependency injection: Custom module-level variable assignment
• In-memory fake services: Configurable classes with controlled behavior
• Property-based testing: Hypothesis library
• Integration testing: Direct calls to AWS services (DynamoDB, EventBridge)
• Automation: Kiro CLI, custom agents, spec-driven development

Business Impact

• Improved developer productivity by reducing the effort required to maintain and evolve the codebase
• Increased confidence in the application's correctness and reliability through comprehensive testing
• Enabled faster iteration and innovation by streamlining the development lifecycle with Kiro's automation capabilities
• Reduced the risk of production issues and customer impact by continuously monitoring and improving the application's quality

Examples

• Circular dependency check in the domain logic, tested using property-based testing with Hypothesis
• Integration with DynamoDB, including validation of error handling and conflict resolution
• Integration with EventBridge, using a test harness to validate the published events