The Agentic Paradigm

Traditionally, testing is focussed on ensuring the pipelines of the factory works correctly.

It does not focus on the quality of content produced.

Logging/tracing, rather than unit and integration tests, are of more value here. It can often be a challenge to create tests for Agentic Systems based on agentic pipelines being different to tradional ones.

Below we can see how tracing/logging is part of the development process that will assist the developer and end user.

We test, evaluate and monitor to ensure our client's needs and wants are met.

Evaluation Driven Development

If we consider our whole app to be a molecule based on a number of atoms, we use EDD to build in evaluations so that development and production can benefit from immediate feedback.