A Black-Box Contract Engine for Agentic Software Development (github.com)

Dojo, a newly announced testing engine built in Go, offers a "black-box" approach for validating software without needing to modify the application code. This innovative tool serves as a transparent proxy, intercepting HTTP traffic and raw database queries to ensure the Software Under Test (SUT) behaves as expected. By embracing a contract-driven testing methodology, Dojo decouples tests from implementation details, preventing traditional unit tests from breaking due to code refactoring by AI agents. This capability is crucial as it reduces the bottleneck often faced during autonomous code generation. The significance of Dojo lies in its ability to streamline testing workflows for complex AI and machine learning applications. By allowing developers to define expected behaviors and outputs clearly through configuration files, while the AI agent handles the coding, Dojo acts as a reliable adjudicator—verifying the correctness of the generated code against defined contracts without being affected by refactoring. Key functionalities include the ability to test various scenarios such as API interactions, regression tests, and chaos engineering scenarios while maintaining compatibility with ongoing migrations across different tech stacks. This makes Dojo a powerful tool for improving the efficiency and effectiveness of AI-driven software development.