Google on Tuesday announced Gemini Code Assist, an AI code completion and assistance tool for enterprises, at its Cloud Next conference.
If this looks familiar, it might be because Google previously offered a similar service under the now-defunct Duet AI brand. The service became generally available in late 2023, but even then Google had already hinted that it would move the service from the Codey model to Gemini in the near future. Code Assist is both a rebrand and a major update to an older service.
Code Assist, which Google Cloud demonstrated at its 30,000-person conference in Las Vegas, will be available through plugins for popular editors like VS Code and JetBrains.
Even more important than the Duet AI version, Code Assist is also a direct competitor to GitHub's Copilot Enterprise and is not the base version of Copilot. That's because there are some Google-specific twists.
This includes support for Gemini 1.5 Pro. Gemini 1.5 Pro is famous for featuring a 1 million token context window, allowing Google's tools to ingest far more context than its competitors. Google says this means, for example, more accurate code suggestions, but also the ability to reason about and change large parts of your code.
“This upgrade delivers a massive context window of 1 million tokens, the largest in the industry. It enables customers to make large-scale changes across their entire codebase, enabling AI capabilities that were previously impossible. Code conversion will be possible with assistance,” Tuesday’s announcement.
Image credit: Google
Like GitHub Enterprise, Code Assist can also be fine-tuned based on your company's internal code base.
“Customizing code using RAG with Gemini Code Assist has significantly improved the quality of Gemini's assistance to developers in terms of code completion and generation,” says Director of Engineering and Head of Generative AI at Turing. Kai Du said. “We expect that introducing code customization will significantly improve overall code acceptance rates.”
This feature is currently in preview.
Image credit: Frederic Lardinois/TechCrunch
Another feature that sets Code Assist apart is its ability to support codebases that are on-premises, such as GitLab, GitHub, and Atlassian's BitBucket, as well as codebases that are split between different services. This is something that Google's most popular competitors in this space currently don't offer.
Google also partners with a number of developer-centric companies to contribute their knowledge base to Gemini. Stack Overflow already announced a partnership with Google Cloud earlier this year. Datadog, Datastax, Elastic, HashiCorp, Neo4j, Pinecone, Redis, Singlestore, and Snyk are also currently working with Google through similar partnerships.
The real test, of course, is how developers react to Code Assist and how useful its suggestions are to them. Google is making the right moves here by supporting a variety of code repositories and providing a large context window, but if the latency is too high or the results are simply not as good, those features None of these are important. And if it doesn't do significantly better than his Copilot, which has a huge head start, it may end up suffering the fate of his CodeWhisperer on AWS, which seems to have almost zero momentum.
It is worth noting that in addition to Code Assist, Google today also announced the launch of CodeGemma, a new open model of the Gemma family fine-tuned for code generation and assistance. CodeGemma is now available through Vertex AI.
Image credit: Frederic Lardinois/TechCrunch
cloud assist
In addition to Code Assist, Google today also announced Gemini Cloud Assist, which helps cloud teams “design, operate, and optimize the application lifecycle.” The tool can, for example, generate architectural configurations tailored to an enterprise's needs based on a description of desired design outcomes. It can also help diagnose problems and identify root causes, and optimize your company's cloud usage to reduce costs and improve performance.
Cloud Assist is available through a chat interface and is built directly into many Google Cloud products.