Artificial intelligence takes over the code review

Tabnine, provider of the code assistant of the same name, announces a new AI agent specifically for code reviews. The system is designed to help improve code quality, security and compliance.

Natural language makes code standards accessible

Development teams can provide the system with their standards in natural language, for example "Never send sensitive information (such as usernames or passwords) in URLs or query parameters, as these are not secure. It is recommended to use headers instead."

Alternatively, users can also provide existing repositories, from which Tabnine's AI then derives specific rules by analyzing patterns and commit annotations. In this way, implicit specifications can also be used for a code review. The rules created by the system are then also available in natural language and can be further adapted, activated or deactivated and weighted by software developers for more control.

In addition to individually definable rules, Tabnine also provides predefined rules that can be activated as required. These include common industry standards as well as best practices specifically tailored to programming languages and frameworks.

Finally, the system automatically checks the code in the development environment and in pull requests, detects deviations and offers suggestions for solutions. As the agent captures the code contextually, i.e. not just as a sequence of characters or syntactic units, there should be fewer false alarms than with conventional review systems for static code analysis.

Private beta for enterprise customers

Tabnine's new code review agent is available to customers on the company's Enterprise plan as a private beta on request. Interested users can also find out more about the documentation for the new system or register for a livestream on November 7, where the company will present the technology.

(mack)