Sentry has been a longstanding aid for developers in monitoring and debugging their production code. Now, the company is enhancing this process by introducing AI Autofix, a new feature leveraging all the contextual data Sentry has on a company’s production environment to suggest fixes whenever an error arises. Despite its name, Autofix is not a fully automated system, recognizing that few developers would feel comfortable with such an approach. Instead, it operates as a human-in-the-loop tool, akin to having a junior developer available for on-demand assistance.
According to Sentry engineering manager Tillman Elser, Autofix is designed to evaluate and address issues from a code-level perspective, rather than focusing solely on system infrastructure performance or errors. Unlike other AI-based coding tools that excel at auto-completing code within an integrated development environment (IDE), Autofix stands out by its ability to proactively identify issues within a company’s production environment. Its primary advantage lies in expediting the process of diagnosing and resolving errors in production, leveraging its understanding of the code’s context. Elser emphasized that Autofix aims to solve production problems swiftly, rather than making developers faster in building applications.
he said; Using an agent-based architecture, Autofix monitors for errors and utilizes its discovery agent to determine if a code alteration could resolve the error. If not, it provides an explanation for why the error cannot be fixed. Crucially, developers are kept informed throughout the process. A notable feature allows developers to provide additional context to the AI agents if they have insights into the issue or simply opt to request a fix directly from the AI.
Once engaged, the AI undertakes several steps to evaluate the issue and devise a plan for resolution. Throughout this process, Autofix offers developers a differential analysis that elucidates the proposed changes. If everything appears satisfactory, Autofix generates a pull request to merge these changes.
