Amazon Q Developer is an AI-assisted software development tool designed to make building software faster and more efficient. The Q Developer Agent helps implement features, fix bugs, and run unit tests based on natural language input.
To initiate the feature development agent, you need to enter ‘/dev’ into the chat field after installing the Amazon Q extension. Its operation is demonstrated using Amazon’s Chronos forecasting model. The agent analyses your codebase to determine how to fulfill requests and produces a detailed plan. If a plan meets your requirements, you can accept it or request revisions. Once a plan is validated, Amazon Q generates the necessary code changes.
Amazon Q Developer Agent uses Artificial Generative Intelligence, which led to it being ranked first on the SWE-bench leaderboard. SWE-bench is a dataset used to test a system’s capability to automatically resolve GitHub issues. Amazon Q’s score was 13.82% on SWE-bench and 20.33% on SWE-bench lite, the highest as of May 2024.
The system overview reveals that upon query submission, a structured representation of a repository’s file system is created. The agent then validates retrieved files and formulates a plan. The content of every retrieved code file is parsed with a syntax parser to obtain a syntax tree of the code. The agent then rewrites sections of code, verifies its syntax, and generates a patch programmatically. This output is shared with you to either accept or request changes.
Public benchmarks like SWE-bench, despite being essential for the AI code generation community, have limitations. The success metric is binary – the code either passes all the tests or it doesn’t. One way Amazon Q adds value is by saving developers time, even when a fully implemented feature isn’t realized instantly.
Amazon Q Developer is free to use with an AWS Builder ID, or through an Amazon Q subscription that offers higher limits. The software was developed by a team of scientists at Amazon Web Services, who continue to improve and evolve it.
