February 8 news.GitHub announced its AI Programming Assistant Copilot A series of major updates and a vision of Copilot's more autonomous future.
One of the most notable features of this update is Copilot's "Vision" feature. This feature allows users to upload screenshots, photos, or diagrams in chat.Based on these images, Copilot generates interfaces, code and alt text to translate the user's visual ideas into actual programmed results.. For example, a member of the marketing team can take a screenshot of a web page and mark it with the changes they wish to make. Whereas previously they would have had to request these changes via text prompts, they can now simply upload the image and ask Copilot directly to make the changes based on the instructions in the file. This feature has been available as an extension in VS Code since last October, and Microsoft confirmed at the time that it would eventually be integrated as a native feature of GitHub Copilot Chat. Previously, users were required to provide their own API key to use it.
GitHub Copilot also introduces "Next Edit Suggestions", an extension of its code completion featuresCopilot. Previously, Copilot performed code completion in the VS Code editor based only on cursor position, but now it can look at other recent edits to predict what developers might want to do next. It automatically recognizes and suggests the next edits through richer context clues. If the developer accepts these suggestions, simply press the Tab key to apply them; if they reject them, press Esc.
1AI notes that last November, GitHub introduced the Copilot Edits feature, which allows developers to make multiple file edits with natural language prompts. Today.This feature is now fully available, with the addition of "proxy mode".In this mode, Copilot automatically recognizes all files associated with a developer's change request without the developer having to manually select the files to apply the change. In this mode, Copilot is able to automatically recognize all files associated with a developer's change request without the developer having to manually select the files to which changes need to be applied. "Copilot is now better able to understand the intent of your initial request and try to resolve it," said Thomas Dohmke, CEO of GitHub. "
In the long run, GitHub wants to apply this "proxy model" to other features of Copilot, enabling broader automation of more interrelated tasks. To that end, GitHub has also revealed a new program called Project Padawan. This is a software engineering agent program that can handle entire tasks independently with guidance from developers. Developers can assign problems to Copilot and let it do the work. However, Dohmke didn't say when the program will officially launch, saying only that GitHub will work with partners and the community to improve it. "We're looking for partners who are also building agents to integrate them into the process, and we want our customers to work with us and provide feedback because we know that at this stage, it's far from perfect."
