GitHub has launched a preview of GitHub Copilot, an AI-based coding assistant for Visual Studio Code that suggests lines of code or functions as you type.
Built in collaboration with OpenAI, GitHub Copilot draws context from the developer’s code, suggesting lines or entire functions while helping to find alternative ways to solve problems, write tests, and explore new APIs without the need to search for answers on the Internet.
Introduced June 29, GitHub Copilot adapts to how the user writes code, helping complete work faster. Trained on billions of lines of public code, the tool is powered by OpenAI Codex, an AI system that is more capable than the GPT-3 (Generative Pretrained Transformer) language model in code generation, GitHub said.
GitHub Copilot can quickly produce boilerplate code and repetitive patterns, with developers able to feed examples to Copilot and have the tool generate the rest. Users also can import a unit test package and have Copilot suggest tests that match implementation code.
GitHub Copilot is available as a Visual Studio Code extension, working wherever Visual Studio Code works, on the developer’s machine or in the cloud on GitHub Codespaces. With access currently limited to a small group of testers, persons interested in trying Copilot can sign up for the Copilot waiting list.
While GitHub Copilot works with a broad set of frameworks and languages, the technical preview works “particularly well” with JavaScript, Python, TypeScript, Ruby, and Go, GitHub said. GitHub cautioned that Copilot does not always write “perfect” code, and that sometimes the suggested code will not work. Code suggested by the tool should be tested, reviewed, and vetted, like any other code, GitHub said.