Integrations are everywhere. Every application integrates with some service somewhere that somebody else wrote. Handling all the possible failure scenarios adds a lot of code to the application, for code we can't even control. How can we test this?

📆 On Tuesday 14 April we invite you to a new session in Tech Talks.

📍Participate physically in the Media Lab or follow the presentation via Zoom.

👾The best integration providers give us access to a test environment. The problem is that these usually only provides for happy paths. We need to be able to make the application fail in all possible and strange ways to ensure we're ready for the real world of unreliable connections.

In this talk Pawel and Martin leverage the code generation capabilities of Claude to create a test server that mimics the provider. This allows them to tailor the responses both for happy paths and all the various failures. They will demonstrate how we can use Skills in Claude to add a test that verifies that the system correctly handles when the exponential back-off algorithm succeeds on the second retry. Skills that describe our architecture and test strategy to Claude allows it to add such test with minimal input from us.

About the speakers:

👨 Pawel Wesolowski: Full stack developer since 2007. Trying to grasp testing and developing test code since 2012. Pawel is an AI-tool enthusiast who paves the road for others.

👨 Martin Larsson: Architect and Developer. Keeping it simple and useful for decades. Martin is more of a cautiously optimistic AI user, focused on creating a safe framework for AI-tools.