Google’s cloud computing division, Google Cloud, has changed the name of its previously released Duet AI for Developers offering to Gemini Code Assist and added new features to make code generation easier.
Gemini Code Assist is based on the Gemini 1.5 Pro model, Google’s most powerful generative AI model. Earlier in December, the company in a blog post had said that it would make its Gemini model available to developers.
Like Duet AI for Developers, Gemini Code Assist provides AI-powered code completion, code generation, and chat. It works in the Google Cloud Console, and integrates into popular code editors such as Visual Studio Code and JetBrains, while also supporting the code base of an enterprise across on-premises, GitHub, GitLab, Bitbucket, or multiple repositories.
The new enhancements and features added to Gemini Code Assist include full codebase awareness, code customization, and enhancements to the tool’s partner ecosystem that increases its efficiency, Google said in a statement.
The full codebase awareness feature, currently in private preview, will allow developers to perform large-scale changes across their entire codebase including adding new features, updating cross-file dependencies, helping with version upgrades, and conducting comprehensive code reviews.
The code customization feature, which is also in private preview, will allow enterprises to connect their private codebases to tailor Gemini Code Assist to help developers in providing context-aware code generation, the company said, adding that this helps the large language model generate more concise and error-free code.
In addition, the company said that it was providing connections for Gemini Code Assist to reach multiple source-code repositories including GitHub, GitLab, and Bitbucket.
In order to increase the efficiency of generating code, the company is expanding Gemini Code Assist’s partner ecosystem by adding partners such as Datadog, Datastax, Elastic, HashiCorp, Neo4j, Pinecone, Redis, Singlestore, Synk, and Stack Overflow.
These partnerships will provide the Gemini model with more data and knowledge from each of these companies to help generate more accurate code.
In December, Google had partnered with Grafana Labs, Confluent, MongoDB, Atlassian, Dynatrace, NetApp, LangChain, and Okta, among others.
Google claims that internal testing of Gemini Code Assist showed that time taken to perform common developer tasks was reduced by 40% and developers spent 55% less time on writing new code.
“We’re also seeing success with customers, like Quantiphi, who recorded developer productivity gains over 30%,” the company said.
Google is not the only cloud service provider that has released code assistance tools underpinned by LLMs. Rivals Microsoft, AWS, and IBM already provide their own offerings in the form of GitHub Copilot, Amazon CodeWhisperer, and IBM Watsonx respectively.