Gemini Leans Into Coding With New GitHub Integration And A Web Tool

gemini-leans-into-coding-with-new-github-integration-and-a-web-tool
Gemini Leans Into Coding With New GitHub Integration And A Web Tool

4

Sign in to your Android Police account

Summary

  • Gemini’s upcoming GitHub extension looks to improve coding efficiency by providing explanations, enhancements, and debugging help.
  • The extension won’t replace general GitHub use, but allows users to ask specific code-related questions and get recommendations.
  • Users over 18 with personal Google accounts will be able to access the tool, which offers a range of code-related assistance in natural language.

While programming might not be the average end user’s idea of AI magic, LLMs like Google Gemini show increasing promise as coding tools. Of course, Gemini only has access to data it’s trained on provided by users, so you won’t find the most recently updated assets from resources like GitHub in its toolkit.

But Google is about to change that, in a way. It will soon roll out a new Gemini extension that lets coders feed the AI a GitHub repository and start firing off requests for code explanations, improvements, and debugging help without looking away from their work. While it has some limitations, it stands to streamline the processes of both learning and producing functional, efficient code, while minimizing time wasted by endlessly slogging through source code (Google via 9to5Google).

What Gemini’s GitHub app can and can’t do

It’s at least as good as a rubber duck

chrome_2018-03-31_19-29-13

Everybody’s favorite software programming resource, Quack Overflow.

For starters, don’t expect the extension to replace navigating GitHub and responsibly maintaining code. You won’t be able to slap a GitHub repository’s URL into a standard prompt and watch the wheels turn. It isn’t slated to have access to metadata like commit history or pull requests, or actually write to a repository. And although it can read private repositories, you’ll need to manually tie your Google account to a GitHub account with the proper permissions before uploading.

But the new development has a ton of potential. By uploading a repository of up to 5,000 files totaling no more than 100MB to a chat, you’ll give the chatbot and its interpreter access to the entire code. Then, you can ask general questions about how the code works, get in-depth explanations of specific functions, and read Gemini’s best efforts at recommending additions, enhancements, and solutions to debugging issues.

Text introducing the Gemini app displayed on a Google Pixel 8

Related

5 best Google Gemini prompts that take full advantage of the generative AI

Ditch boring emails and tedious tasks; these Gemini prompts will transform how you work

Here are some examples of what Gemini can handily answer and achieve with a provided codebase:

  • What external libraries are used in the attached code?
  • How can the attached code be more efficient?
  • Write event handlers for the attached code to recommend products to users.
See also  Google Launches Gemini 2.0 Pro And Flash-Lite With Improved Reasoning And Efficiency

It’s a subtly exciting addition to the ever-improving slate of tasks Gemini’s explicitly prepared to take on. It hasn’t launched just yet, but it will be available to users over 18 years old who are signed in to personal Google accounts and have Gemini Apps activity turned on. Like the helpful Gemini Code Assist feature, it won’t be restricted to paid subscriptions, at least not in its first form.