Revolutionize Your Workflow with Gemini Code Assist

Author(s): Dimitre Oliveira

Originally published on Towards AI.

In the current fast-paced and evolving world of software development, staying ahead of the curve is crucial. AI-powered tools are no longer a futuristic concept but a present-day reality that can significantly boost productivity and code quality. For this, Google’s Gemini Code Assist is a very good option. It offers a suite of features designed to improve the entire development lifecycle. This blog post will be your guide to learn and getting started with Gemini Code Assist, focusing on three key areas: the GitHub plugin for code reviews and summaries, integrating the IDE plugin into your daily coding workflow, and Agent Mode for complex tasks.

Effortless Code Reviews with the Gemini Code Assist GitHub Plugin

Good code reviews are mandatory for any successful development team, ensuring code quality, consistency, and knowledge sharing. However, they can also be a time-consuming process. Gemini Code Assist for GitHub transforms this critical practice by acting as an AI-powered code reviewer, automatically summarizing pull requests and providing in-depth feedback.

Here are the steps required to set up and start using Gemini Code Assist for GitHub

1. Go to the Gemini Code Assist app page.
2. Sign in to your GitHub account if you haven’t already.
3. Click Install.
   A prompt to install the Gemini Code Assist app for a user or organization is displayed.
4. When prompted to install the Gemini Code Assist app for a user or organization, select the organization you intend to use it on.
   After you’ve installed the Gemini Code Assist app for your GitHub organization, you’re prompted to select the repositories to enable the Code Review integration.
   You’re redirected to the Admin Console for the Gemini Code Assist app.
5. Login with your GitHub account.
6. Select a GitHub organization or personal account from the drop-down menu.
7. Review and accept the Google Terms of Service, Generative AI Prohibited Use Policy and Privacy Policy, and then click Complete setup.
   Gemini Code Assist is added to the pull requests within your selected repositories.

Once you’ve set up the Gemini Code Assist for GitHub, it will automatically provide an initial review when a new pull request is opened. This review includes a summary of the changes and comments on specific parts of the code, with severity levels ranging from “Critical” to “Low”. The feedback covers a wide range of areas, including correctness, efficiency, maintainability, and security.

Gemini Code Assist also allows customization. You can tailor Gemini’s behavior to your team’s specific needs by adding a .gemini/ folder to the root of your repository. Inside this folder, you can create two key files:

• config.yaml: This file allows you to configure various settings, such as disabling reviews, setting a severity threshold for comments, and specifying files to ignore using glob patterns.
• styleguide.md: Here, you can provide a natural language description of your team’s coding conventions and best practices. Gemini will then use this style guide to ensure that all code reviews align with your established standards.

You can also manually interact with Gemini in a pull request by using commands like /gemini summary to get a summary of the changes or /gemini review to request a code review.

Supercharge Your IDE with the Gemini Code Assist Plugin

Gemini Code Assist seamlessly integrates with your favorite IDEs, including Visual Studio Code and JetBrains IDEs, bringing the power of AI directly into your coding environment. This integration provides a host of features designed to make your day-to-day workflow smoother and more intuitive.

To use Gemini Code Assist on any IDE, you must follow the setup steps, which usually consist of searching for the extension and installing it (guides for enterprise and individuals).

The IDE plugin offers a chat interface that is aware of your code, providing context-specific assistance and answering your questions without the need to switch between different applications. It also provides automatic code completions as you type, and can even generate or transform entire functions or files on demand.

Recent updates have further enhanced the IDE experience. You now have more granular control over the context that Gemini uses, with the ability to use a .aiexlude file to ignore sensitive or legacy code. You can also attach terminal output directly to the chat to get help with commands or debug errors, eliminating the need to copy and paste logs. These refinements help you stay in a state of flow and focus on what matters most: solving problems.

Unleash the Power of Agent Mode

For more complex tasks, Gemini Code Assist has a feature called “Agent Mode”. This mode transforms the AI assistant into an AI pair programmer that can analyze your entire codebase to plan and execute complex, multi-file tasks from a single prompt.

With agent mode, you can do any of the following and more:

• Ask questions about your code.
• Use context and built-in tools to improve generated content.
• Configure MCP servers to extend the agent’s abilities.
• Get solutions to complex tasks with multiple steps.
• Generate code from design documents, issues, and TODO comments.
• Control the agent behavior by commenting on, editing, and approving plans and tool use during execution.

Here is a quick summary of how it works:

In agent mode, your prompt is sent to the Gemini API with a list of tools that are available. The Gemini API processes the prompt and returns a response. The response might be a direct answer or a request to use an available tool.

Agent Mode is capable of handling a wide range of tasks, from implementing new features and performing large-scale refactors to upgrading dependencies and fixing breaking changes. It works by creating a detailed plan of the files to be modified and the changes to be made, which you can then review and approve before any code is altered.

A great example of Agent Mode in action is its ability to generate a README.md file for an existing application. By simply providing the prompt “Generate a README.md file for this application”, the agent will analyze the repository and create a comprehensive README that explains the application’s functionalities, structure, and how to run it locally.

You can also use Agent Mode to implement new features. For instance, if you want to split a single larger function into separate functions, you can provide a high-level prompt describing the desired outcome. The agent will then devise a plan to modify the necessary files, including creating new routes, forms, and backend logic.

Furthermore, Agent Mode can be used to generate and run unit tests. With a prompt like “Generate unit tests for all the application functions and run them to verify that everything works as expected,” the agent will create the necessary test files and execute them. If any errors are encountered, it will attempt to find the cause and implement the required fixes autonomously.

To customize the agent’s behavior, you can create a GEMINI.md file in your project’s root directory. This file can contain rules, style guides, or any other context that you want the agent to consider when performing its tasks. It is also possible to configure tools for agent mode, from the built-in tools like grep and file read or write, local or remote Model Context Protocol (MCP) servers.

Conclusion

Gemini Code Assist is more than just a coding assistant; it’s a comprehensive tool that can revolutionize the way you and your team work. From simplifying code reviews with the GitHub plugin to supercharging your IDE and tackling complex tasks with Agent Mode, Gemini empowers you to write better code faster. Whether you’re an individual developer or part of a large enterprise, Gemini Code Assist is a valuable addition to your development toolkit.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI