Ushering in a new era for app developers

Generative AI is transforming software development with tools that automate tasks, foster collaboration, and spark innovation. Its ability to boost creativity and productivity for developers is undeniable. We’re empowering developers with products to maximize generative AI’s potential, both in streamlining their work and building their own AI-powered applications. 

Gemini Code Assist, the evolution of Duet AI for Developers, supercharges your development process while improving code quality and security. With new additions rolling out in Gemini Code Assist, now you can get:

• Full codebase awareness: Unlock the full potential of your codebase with Gemini Code Assist’s full codebase awareness feature, now available in private preview. Make complex changes, add groundbreaking features, and streamline updates with deep understanding using Gemini 1.5 Pro, Google’s state-of-the-art model with the world’s largest context window. 

• Transform your code effortlessly: With the new code transformation feature, available today in Cloud Workstations and Cloud Shell Editor, you can now use natural language prompts to tell Gemini Code Assist to analyze, refactor, and optimize your code. 

• Extended local context: To support tasks requiring information across multiple code files with higher accuracy, Gemini Code Assist now automatically retrieves local files from your IDE workspace that are relevant, and displays references to the files used.

• Code customization: Integrate your private codebases, and repositories for hyper-personalized (private preview) code generation and completions powered by our Gemini models. Gemini Code Assist now connects to multiple source-code repositories, including GitLab, GitHub, and Bitbucket.

• Integration services: We are expanding the reach of Gemini Code Assist with Apigee and Application Integration to access and connect your applications.

• New partnerships: Following our ecosystem announcement last December, we’re partnering with Snyk to bring security knowledge to your fingertips. Gemini Code Assist, powered by Snyk’s security insights, lets you learn about vulnerabilities and common security topics right within your integrated development environment (IDE) for instant, context-aware guidance.

Get full codebase awareness 

With Gemini Code Assist’s full codebase awareness feature, currently in private preview, you can perform large-scale changes across your entire codebase, including adding new functionality, updating cross-file dependencies, helping with version upgrades, comprehensive code reviews, and more. Gemini Code Assist’s massive one-million-token context window and full code awareness revolutionize coding. It understands your entire repository, not just individual files, providing insights that span over 100,000 lines of code. 

Boosting developer productivity with code transformation 

Gemini Code Assist’s code transformation feature lets you leverage natural language to analyze, refactor, and optimize existing code directly within your IDE. Here is just one example of code transformation in action. In the image below, you can see that I highlight the function and clear, descriptive comments are generated on-demand in my IDE.

Speaking of IDEs, a key way we’re working to enhance the developer experience is with tools that integrate into existing developer workflows. Gemini Code Assist supports popular IDEs like Visual Studio Code and the full suite from JetBrains. In fact, you can now leverage the productivity enhancements of Gemini Code Assist across Google Cloud. For instance, Gemini Code Assist in Apigee, now available in public preview, empowers you to effortlessly create enterprise-grade APIs by generating specifications, mocking behavior, and building proxies – all with natural language prompts. Likewise, Gemini Code Assist in Application Integration, now available in public preview, suggests new flows or extensions, intelligently reusing your assets for maximum efficiency.

Code customization with your private codebase

Gemini Code Assist’s new code customization feature is now available in private preview, which allows users to leverage their private codebases for more relevant responses, or instruct Gemini models to guide them toward existing solutions within their organization’s code repositories. 

One company who is using code customization to complete projects faster, more efficiently and with better quality is Turing, an AI-powered tech services company that helps clients train large language models (LLMs) and build gen AI solutions. The results speak for themselves: with Gemini Code Assist code customization, Turing found acceptance rates improved by nearly 70% compared to the base model alone.

Here’s how Turing further benefits from code customization with Gemini Code Assist:

Gemini Code Assist works within the context of Turing’s specific codebase. This means less time spent through documentation and more time on coding. Gemini Code Assist understands Turing’s internal code repositories, accurately suggesting the right functions (like ‘getFeatureList()’) and even incorporating relevant ‘experimentationSDK’ references. The result: faster development of correct, executable code.”

“Turing’s Intelligent Talent Cloud helps companies build and scale their engineering teams with a global pool of top remote developers. Gemini Code Assist empowers our developers to write more correct code the first time, resulting in productivity gains of 33% and demonstrating how Gemini Code Assist can accelerate development on the Turing platform.” – Kai Du, Director of Engineering, Turing

As with everything that we do at Google, security and compliance is paramount. We’re building robust features into code customization like single-tenancy, isolated VPC, and enterprise access control, along with an auditable access history, to help ensure your private code is secure. We also recognize that source code lives in various locations. This is why Gemini Code Assist with code customization connects to popular source code repositories like GitHub, GitLab, and Bitbucket. We also support various deployment models — private, self-hosted, and on-premises.

An expanded partner ecosystem

Last year we announced that our industry-leading ISV partners will allow us to enhance Gemini Code Assist using their data, integrating documentation, knowledge sources, and other signals to optimize Gemini Code Assist for more contextualized assistance about their products. This allows partners like Datadog, DataStax, Elastic, HashiCorp, MongoDB, Neo4j, Pinecone, Redis, SingleStore, and Snyk to improve Gemini Code Assist responses, bringing greater depth of intelligence for their products.

Today we are announcing an expansion of our partnership with Snyk, a leader in developer security tools, to help developers build applications quickly and securely across the software development lifecycle. Now you can get actionable security guidance by Synk directly within your workflow. Ask Gemini Code Assist how to use Snyk, address common vulnerabilities, and get best practices – all backed by Snyk’s security expertise. For example, ask “How can I enforce Snyk scans before my code gets deployed to GKE?” and Gemini Code Assist will provide detailed instructions. Or, query “What is Snyk’s suggestion on using Hard Coded Secrets?” and receive risk analysis and remediation advice.

Available later this year, Gemini Code Assist will offer a powerful new security capability with Snyk: the ability to proactively scan code snippets directly within your IDE. 

“Snyk is thrilled to announce our newest collaboration with Google Cloud to create a more seamless security experience for developers leveraging Gemini Code Assist. The integration will allow developers to easily identify and fix vulnerabilities in human- or AI-generated code — accelerating software development initiatives, securely.” – Danny Allen, CTO of Snyk

Making gen AI application development easier

Creating a gen AI app may seem deceptively simple. However, developing a robust, enterprise-grade solution requires far more than just selecting a model and tweaking your prompts. You also need to proactively reduce the hallucinations these models sometimes produce, leveraging techniques such as Retrieval Augmented Generation (RAG), model tuning, factual grounding, and MLOps – a substantial investment. 

This speaks to Google Cloud’s mission: to reduce friction and help developers build production-grade gen AI applications with speed and ease. We want to be your partner in this endeavor, and we’re supporting developers in three key ways.

Get started quickly when learning and evaluating new technologies
We provide tools and resources to help you jumpstart your gen AI projects, enabling you to move from concept to working prototype quickly and efficiently. One powerful resource is our collection of gen AI-focused Jump Start Solutions, available directly within the Google Cloud console. For example, our solution for building a RAG application with Cloud SQL empowers you to deploy a gen AI app with a few clicks. Jump Start Solutions provide a hands-on way to accelerate your understanding and implementation of powerful gen AI concepts. Once you’re ready to test and refine your ideas, Google AI Studio provides the prototyping environment where you can experiment with gen AI concepts quickly

An expansive portfolio for building and running gen AI apps
We offer an expansive set of easy-to-use yet flexible tools to build and run your gen AI apps. For starters, Vertex AI Model Garden’s curated library of 130+ models (with easy discovery, customization, and deployment tools) gets you started fast. Database solutions like AlloyDB integrate enhanced AI capabilities specifically for RAG-based gen AI apps, helping to ensure your users always have the most relevant, up-to-date data. Finally, our scalable GKE and Cloud Run container platforms provide the efficiency and flexibility you need to run your gen AI apps at scale. 

Meeting you where you are with popular ecosystem technologies
We’re constantly expanding our integrations with our tools to streamline your gen AI development workflow. This includes deep integrations with popular open-source frameworks like LangChain, with our robust database offerings like AlloyDB, Cloud SQL, and Spanner. For easy scaling and a developer-friendly experience, Cloud Run now supports LangChain applications. We’ve also partnered with Hugging Face, a vibrant AI community, to empower developers even further. Our unique integration enables one-click model deployment from Hugging Face to Vertex AI or GKE, allowing Hugging Face users to leverage Google Cloud’s TPU power for faster training and deployment. It’s no wonder that over 90% of AI unicorns choose to build on Google Cloud.

Ready to create your own gen AI shopping assistant and put your skills to test? Get started with our easy RAG demo on GitHub. All of our exciting announcements today, including full codebase awareness, code transformation features, code customization, and new partnerships, are all designed to empower you, the developers. We’re making it faster and easier than ever to build cutting-edge generative AI applications. Join 5,000+ organizations transforming their workflow with Gemini Code Assist. Experience the power first hand by signing up for our preview program.