AWS has added features to its CodeWhisperer AI assistant for developers, including support for infrastructure as code (IaC), security scanning and a preview of support for C# in Visual Studio, adding to the recent introduction of command-line AI on macOS.
Developers can now get AI-driven authoring suggestions when typing or editing code for AWS Cloud Formation, AWS CDK (Cloud Development Framework), or HashiCorp Terraform HCL (HashiCorp Configuration Language).
Security scanning was always part of CodeWhisperer, but previously only for detection and only for Java, Python or JavaScript. The new update adds support for TypeScript, C#, and the new IaC languages. It now also provides code suggestions for “identified security and code quality issues,” according to the company’s post, which comes just as the annual re:Invent shindig gets under way in Las Vegas.
AWS is also previewing support for C# in Visual Studio, adding to existing support for Visual Studio Code (a very different thing), JetBrains IDEs, JupyterLab, AWS Glue Studio Notebook, Cloud9, Lambda console, Glue Studio and SageMaker Studio. The support for C# only is odd in some ways, since CodeWhisperer is more oriented towards JavaScript and Python, but makes sense as part of an AWS strategy to win .NET developers to its cloud platform.
A week or so ago, AWS also added command-line support on macOS only, including in VS Code and JetBrains IDEs. In the command line, CodeWhisperer provides code completions for more than 500 command-line interfaces, such as for git, npm and Docker, as well as natural language to bash translation.
The example given for natural language translation is “reverse my most recent git commit”, which will be translated to an executable shell code snippet. Natural language translation is triggered by typing q followed by a prompt.
Shell support requires a separate download and uses accessibility support on the Mac, which means giving the AWS tool a high level of permissions – off-putting for some. It has the potential to be useful because remembering the right CLI incantation for a task can be challenging; though like all generative AI tools it is not completely reliable and at worst could execute a wrong or dangerous command. One of the options is not to allow “instant execute of dangerous suggestions.”
The activity around CodeWhisperer suggests that AWS is investing substantial resources into the product, which lacks developer mindshare and usage compared to rivals such as GitHub Copilot. Last month AWS introduced the ability to customize CodeWhisperer with enterprise internal code and documents, an important feature which was added ahead of Copilot’s similar one. Another attractive feature of CodeWhisperer is that it remains free to individual developers, though with limitations such as only 50 code security scans per month. A professional license costs $19 per user/month. A distinctive aspect of CodeWhisperer is that it is trained on “billions of lines of code, including Amazon and open source code,” according to the FAQ, which makes it highly suitable for working with AWS services but perhaps less suitable for developers not using the Amazon platform – which is hardly surprising.