Google unveiled Gemini Code Assist, a potent tool that offers enterprise users code completion and assistance, during the Cloud Next conference.
If this sounds familiar, it’s probably because Google previously provided a similar service under the now-defunct Duet AI branding. Google made the service widely accessible in late 2023, indicating its intention to transition from the Codey model to Gemini in the near future. Code Assist is a new and improved version of the previous service.
Google Cloud showcased Code Assist at its large conference in Las Vegas. It will soon be accessible through plug-ins for widely used editors such as VS Code and JetBrains.
Code Assist is a strong contender against GitHub’s Copilot Enterprise, rather than the basic version of Copilot. In terms of capabilities, it surpasses even the Duet AI version. That’s because some Google-specific features are unique.
One of the notable features is the support for Gemini 1.5 Pro, known for its extensive million-token context window. This unique capability sets Google’s tool apart from its competitors, enabling it to gather a significantly larger amount of contextual information. This results in more precise code suggestions, as well as the ability to analyze and modify substantial portions of code.
This upgrade introduces an impressive 1 million-token context window, setting a new industry standard. Customers now have the ability to make extensive modifications to their entire code base, opening up new possibilities for AI-powered code transformations. Brad Calder, Google’s VP and GM for its cloud platform and technical infrastructure, shared this exciting news during a press conference prior to Tuesday’s announcement.
Code Assist, like GitHub Enterprise, allows customization to match a company’s internal code base.
Kai Du, the Director of Engineering and Head of Generative AI at Turing, stated that the use of RAG with Gemini Code Assist significantly improved the level of support Gemini provided to our developers, particularly in terms of code completion and generation. With the implementation of code customization, there is anticipation of a significant rise in the overall code acceptance rate.
This feature is currently in preview.
Code Assist offers exceptional support for codebases located on-premises or on popular code hosting platforms like GitLab, GitHub, and Atlassian’s BitBucket. It can effortlessly manage codebases dispersed across multiple services. Google’s top rivals in this field currently do not provide that feature.
Google is collaborating with several companies that focus on developers to integrate their knowledge bases into Gemini. Earlier this year, Stack Overflow announced its partnership with Google Cloud. Datadog, Datastax, Elastic, HashiCorp, Neo4j, Pinecone, Redis, Singlestore, and Snyk have recently formed partnerships with Google, expanding their collaborative efforts.
The true measure, naturally, lies in the response of developers to Code Assist and the practicality of its suggestions for them. Google’s strategic decisions to support multiple code repositories and provide an extensive context window demonstrate their expertise in artificial intelligence. However, it is crucial for them to address any issues related to latency and ensure that the search results are of high quality, as these factors ultimately determine the effectiveness of their features. If the new system fails to surpass Copilot, despite Copilot’s early advantage, it could face a similar fate as AWS’ CodeWhisperer, which appears to have little to no momentum.
It is important to mention that alongside Code Assist, Google has also introduced CodeGemma, a new open model in its Gemma family specifically designed for code generation and assistance. Vertex AI now provides access to CodeGemma.
Cloud Assist
Google has introduced Gemini Cloud Assist, a tool that aims to support cloud teams in effectively managing their application lifecycle. This new offering complements the existing Code Assist feature. A company’s specific requirements, as outlined in a description of the desired design outcome, can guide the tool’s generation of customized architecture configurations. Furthermore, it can assist in identifying problems and determining their underlying causes, while also enhancing a company’s cloud utilization to minimize expenses or increase efficiency.
Cloud Assist will be accessible via a chat interface and seamlessly integrated into various Google Cloud products.