Code Customization with Gemini Code Assist

Quiz

Sometimes, you need to think out of the box to generate exact code suggestions from Gemini Code Assist. Where it supports multiple languages and frameworks, the magic happens when you master the customisation of code suggestions. With the help of this, Gemini understands your whole project which is linked and then it suggests all the code suggestions accordingly, which will be helpful in future coding.

Here, we will see steps to link your project with Gemini Code Assist to get you the best-customised output. This customisation ensures that Gemini understands the unique patterns and coding standards of your team.

Understanding Code Customisation in Gemini

Code customisation is the process where Gemini Code Assist analyses your projects linked repository to offer more precise and advanced code suggestions. Instead of providing generic solutions, Gemini understands the libraries, APIs, and dependencies within the repository and adapts its recommendations accordingly.

As a result, developers receive suggestions that not only match their framework but also give out the code which suits the best practices and structure specific to the project.

This process eliminates the need for manually recalling custom functions, services, or coding conventions, allowing you to focus on development rather than boilerplate code.

The Architecture of Code Customisation in Gemini

Lets discuss the architecture of code customization in Gemini −

Normally, Gemini Code Assist works with the support of cloud services provided by Google, like different cloud workstations, and cloud shell editors and integrates it with IDEs like VSCode.

Gemini provides normal solutions, based upon the dataset it is trained on, giving out the most common troubleshoots and code suggestions which seems okay, but it does not get personalised with your current projects.
To achieve better code suggestions, you can link your repository with your working environment and get better results.

This is how Gemini works when you connect your private code repository with your IDEs environment where Gemini works.
For this, a tool helps you out, which is also called the Developer Connect, which connects your private repository with Gemini services.
The repository can be a GitHub repository, GitLab repository, or any of the one which is supported by Developer Connect.

When Gemini Code Assist links with your repository, it follows a three-step architecture to deliver personalised suggestions −

Step 1: Repository Analysis

Gemini scans the repository files, libraries, and dependencies.
It identifies patterns like naming conventions, APIs, and modules used within the codebase.

Step 2: Contextual Code Modelling

Gemini builds a customised code model based on past code usage, imports, and your projects structure.
This model helps the AI anticipate the next steps in your development process.

Step 3: Real-Time Suggestions

As you write code, Gemini provides project-specific recommendations.
It intelligently completes functions or suggests optimised solutions by leveraging the linked repositorys context.

There are several programming languages which support this optimization, covering almost all kinds of popular languages and frameworks. Some of the optimised programming languages are −

C / C++
C#
Golang
Python
Java
JavaScript
Kotlin
PHP
Rust
Typescript

Steps to Customise Code Suggestions in Gemini

Heres a step-by-step guide on how to customise code suggestions through repository linking −

Step 1: Connect Your Repository

Link your GitHub, GitLab, or Bitbucket repository with Gemini Code Assist through the Gemini dashboard.
Go to the configure Gemini settings in the IDE which you are using.

After that, you will get options to link your private repository (like Github Repo) to your working environment.
Ensure that the repository contains your projects dependencies and essential code modules.

Step 2: Analyse Dependencies Automatically

Once linked, Gemini scans the libraries, APIs, and code patterns within the project.
No manual setup is needed, Gemini builds a contextual code model.

Step 3: Set Coding Preferences (Optional)

Adjust settings for naming conventions, indentation, or code style to align suggestions with team standards.
Example: Specify naming conventions like camelCase for variables or PascalCase for classes.

Step 4: Iterate and Improve

As you commit changes to the repository, Gemini continuously learns and updates its suggestions.

Example of Code Suggestions

Lets take some programming languages with frameworks to show examples of code suggestions −

In Python + Django

Before repository linking, Gemini can provide basic Django suggestions like standard views or models. However, once your repository is linked, the suggestions become more project-aware, adapting to the APIs and patterns already in use.

Before Repository Linking −

# Basic view suggestion
def sample_view(request):
   return HttpResponse("Hello, World!")

After Repository Linking −

from myproject.serializers import CustomSerializer

def sample_view(request):
   data = CustomSerializer(data=request.data)
   if data.is_valid():
      return JsonResponse(data.data)
   return JsonResponse(data.errors, status=400)

With the repository linked, Gemini recognizes that the project uses CustomSerializer and suggests it automatically. This eliminates the need to search for existing code snippets manually, speeding up development.

Java + Spring Boot

In a Java + Spring Boot project, Gemini offers general suggestions but evolves with advanced recommendations after linking your repository.

Before Repository Linking −

public class HelloController {
   @GetMapping("/hello")
   public String sayHello() {
      return "Hello, World!";
   }
}

After Repository Linkin −

@RestController
@RequestMapping("/api")
public class UserController {

   @Autowired
   private UserService userService;

   @GetMapping("/user/{id}")
   public ResponseEntity<User> getUser(@PathVariable Long id) {
      return ResponseEntity.ok(userService.getUserById(id));
   }
}

Here, Gemini has learned about your project repository and found out that your project uses UserService and suggests injecting it directly into the controller. It aligns the generated code with your project's structure, avoiding redundant code.

JavaScript + Node.js

For JavaScript projects, Gemini can provide intelligent code completions for frameworks like Node.js. Below is an example of how repository linking can upgrade suggestions.

Before Repository Linking −

// Basic route suggestion
app.get('/hello', (req, res) => {
   res.send('Hello, World!');
});

After Repository Linking −

// Here is the advanced suggestion having custom middleware and service
const { validateUser } = require('./middlewares/validateUser');
const { userService } = require('./services/userService');

app.get('/user/:id', validateUser, async (req, res) => {
   try {
      const user = await userService.getUserById(req.params.id);
      res.json(user);
   } catch (error) {
      res.status(500).send(error.message);
   }
});

Here, the validateUser middleware and userService call, ensure your code follows the existing project patterns. This improves code quality and reduces development time.

Benefits of Custom Code Suggestions in Gemini

The following are the benefits of custom code suggestions in Gemini −

Accelerates Development
Aligns with Team Standards
Minimises Errors and Bugs
Integrates Seamlessly with CI/CD
Improves Code Readability and Maintainability
Reduces Onboarding Time for New Developers
Enhances Productivity with Real-Time Feedback
Supports Seamless Collaboration Across Teams
Integrates Smoothly with Version Control Systems
Keeps Code Consistent Across Multiple Frameworks and Languages

Limitations and Challenges

Lets discuss some limitations and challenges of customizing code suggestion using Gemini Code Assist −

Dependency on Repository Quality

If the linked repository is poorly structured or lacks proper documentation, Geminis suggestions may be less effective.

Limited Version Capability

Here, the code customisation is supported only in the VS Code with Gemini Assist + Cloud Code extension for versions v2.18.0 and above. IN IntelliJ with its cloud code plugin, Gemini code customisations are supported in version 1.1.0 and above.

Limited Linking

This feature is under development and Google provides only one code repository index which can be linked for each project. You need a higher quota for more code repository indexes. You can get more info here.

Framework Complexity

Projects using multiple frameworks might lead to less precise recommendations.

VPC accessibility

Code customisation does not allow for Virtual Private Cloud accessibility, limiting data theft recovery and increasing accidental data loss threats.

Region Limitations

Currently, Geminis code customisation works with the Developer Connect connections in regions like asia-southeast1, us-central1 and europe-west1 only.

Conclusion

Linking your repository not only makes the suggestions more relevant but also ensures better code quality and consistency. Hence you can always focus on the creativity and problem-solving part, not worrying about the structural approach.

Print Page