GitHub has announced major updates to its AI-powered coding assistant, Copilot. The latest upgrades include adding the ability of Copilot's "agent mode" to add to its own code to detect and fix errors.
The agent mode in Copilot supports terminal command suggestions and runtime error analysis, among other features. It can also execute additional code not requested but necessary to make the main request work.
To get started, you’ll need to download VS Code Insiders and then enable the agent mode setting for GitHub Copilot Chat. Then, when in the Copilot Edits panel, switch from Edit to Agent right next to the model picker. Agent mode will change the way developers work in their editor; and as such, we will bring it to all IDEs that Copilot supports.
Alongside the agent mode, GitHub also announced the general availability of Copilot Edits in Visual Studio Code. Combining the best of Copilot's chat and inline capabilities, developers can indicate a set of files to edit and then instruct Copilot with natural language to make the desired changes.
GitHub adds that Copilot Edits can make inline changes across multiple files, presenting the user with a conversational flow and giving them the ability to review and accept suggested modifications. Under the hood, Copilot Edits leverages a dual-model architecture. There's a base language model taking into account the full context of the editing session to make the initial edit suggestions.
Users can choose which foundation model to use, including GPT-4o, o1, o3-mini, Claude 3.5 Sonnet, and Gemini 2.0 Flash. The suggested edits are then dispatched to a speculative decoding endpoint optimized for fast application of changes in the files. They can also run the code to verify the changes and, where necessary, revert the suggested modifications.
Copilot Edits is in the Secondary Side Bar (default on the right) so that you can interact with views in the Primary Side Bar, such as the Explorer, Debug, or Source Control view, while you’re reviewing proposed changes.
The GitHub roadmap for Copilot Edits includes improving performance for the apply changes speculative decoding endpoint, allowing transition into Copilot Edits from Copilot Chat without loss of context, providing a way to suggest files to the working set, and finally, providing an option to undo suggested chunks.
1 Comment - Add comment