Google Cloud, the cloud division of Google LLC, today announced that generative artificial intelligence assistance is coming to software developers where they do their work, in developer environments and editors, with Duet AI for Developers.

The launch of Duet AI for Developers into general availability marks the technology giant’s first push into the AI-powered coding assistant space with a competitor to Amazon.com Inc.’s CodeWhisperer and GitHub Inc.’s Copilot. It joins Duet AI for Workspace, which has been in general availability since August. This offering is related to a generative AI program first announced by Google in May as part of Duet AI for Cloud during an early testing release.

“Duet AI for Developers is this first foray that helps developers and site reliability engineers — it’s not just for devs,” Richard Seroter, chief evangelist at Google Cloud, said in a press briefing. “We are also trying to make sure that operators have some assistance with troubleshooting and best practices.”

The Duet AI assistant is available across multiple integrated development environments, which are applications that developers use to write, test, compile and deploy code. The IDEs it can integrate with include Cloud Shell Editor, Cloud Workstations, IntelliJ, PyCharm and Visual Studio Code. Duet supports more than 20 programming languages, including C, C++, Go, Java, JavaScript and Python.

Seroter explained that the assistant provides everything a developer might want out of an AI-coding companion with a chat interface and code completion. It also helps reduce context switching by providing a way to deliver knowledge directly to the programmer within their development environment so that they don’t need to exit their flow while writing code. They can just move over to the chat with the AI assistant, ask a question or request code, and then go directly back to coding.

It also stays out of the way of the developer and lets them handle the coding. “This is fully managed, you don’t think about installing things,” said Seroter. “It’s just fully activated in the cloud for the user.”

Aside from generating code, the AI assistant will additionally be able to explain code to developers, which will come in handy if a project has just been handed to them. It can also produce test cases directly out of code, simply right-click on a segment of code and ask the AI to generate a test.

In the coming weeks, all of Duet AI services will be incorporating Gemini, Google’s recently announced biggest and most capable AI model, which includes advanced reasoning and complex coding ability.

In addition to these capabilities, Duet customers will be able to customize their models with their own codebases to receive specialized code suggestions that are informed by their own private code. This will mean that the suggestions and code blocks provided will stay in line with conventions and practices unique to the enterprise. This capacity is currently in testing and is planned to roll out sometime in the second quarter of 2024.

Seroter said that Google is currently working with a growing ecosystem of partners that will contribute datasets for code-assist and knowledge bases for Duet AI for Developers to enhance its capabilities. Using knowledge from these partners, AI assistance from Duet will be able to provide additional assistance based on partners’ coding and data models, product documentation, best practices and other resources.

Code-assist partners such as Confluent Inc., HashiCorp Inc. and MongoDB Inc. provided code that will allow Duet to generate code specific to those platforms. This will allow developers to build and troubleshoot faster using their own tools. Knowledge base partners such as Datadog Inc., JetBrains and LangChain have offered documentation and partner-specific information about how to resolve common production issues and how to automate tests for quality assurance.

Starting today, Duet AI for Developers is available for no cost until Feb. 1, 2024, after which it will become a subscription service. Software engineers and operations teams can try it out.

Photo: Pixabay