GitHub Copilot Chat is coming to GitHub

Pledging to being “re-founded” on its budding Copilot AI coding technology, GitHub on November 8 announced general availability plans for GitHub Copilot Chat and previews of GitHub Copilot Enterprise as well as new AI-powered security capabilities.

GitHub Copilot Chat, which has been in beta stage, will be generally available in December as part of existing GitHub Copilot subscription programs. It provides a chat interface to interact with GitHub Copilot, using whatever language a developer speaks. Answers to coding-related questions can be received from within a supported IDE, including JetBrains IDEs, Microsoft’s Visual Studio Code and Visual Studio, and the Neovim editor. Support for JetBrains is available in preview now.

GitHub Copilot Chat is powered by OpenAI’s GPT-4 large multimodal model, bringing more accurate code suggestions, explanations, and guidance, GitHub said. An inline capability in GitHub Copilot Chat helps developers chat about specific lines of code. GitHub also is introducing slash commands to streamline tasks such as creating unit tests.

GitHub also is integrating GitHub Copilot Chat directly into the GitHub platform at github.com, so developers can dig into code, pull requests, documentation, and general coding questions. GitHub Copilot Chat also will be available in GitHub’s mobile app as part of GitHub Copilot subscriptions. Developers can get answers to programming questions by typing into iPhone an Android devices or use speech-to-text capabilities in those devices.

GitHub Copilot Enterprise allows organizations to personalize GitHub Copilot with the full context of a code base. With GitHub Copilot Chat connected to repositories on github.com, Copilot Enterprise lets teams search and build documentation, get suggestions based on internal and private code, and review pull requests. Also encompassing the GitHub Copilot for Business program, GitHub Copilot Enterprise is due in February 2024 for $39 per user per month.

Trained on natural language text and source code from publicly available sources, GitHub Copilot has not been without controversy. Questions have arisen over the legality of its use of open-source licensed code for training. A class action lawsuit filed a year ago is still in the process of going through the courts and there has been no settlement, according to GitHub. The company expressed confidence that GitHub Copilot adheres to applicable laws.

GitHub also made these announcements on November 8:

• The GitHub Copilot Partner Program will create a plugin ecosystem for GitHub Copilot, with the goal of broadening the expanse of what developers can do with AI. GitHub envisions use cases such as helping to improve the performance of database queries, checking the status of a feature flag, and viewing results of an A/B test.
• The research team at GitHub Next has developed an AI-powered bridge to help developers scale the barrier of putting ideas into code. Called the GitHub Copilot Workspace, due in 2024, this technology will present a plan for implementing software changes, which then are built and tested and validated for success. If developers introduce an error, it will be patched and the code rerun.
• GitHub Advanced Security is being enhanced with AI-powered application security testing designed to detect and remediate vulnerabilities and secrets in code. Subscriptions to the service soon will include features in preview such as code scanning autofix, which suggests AI-generated fixes for JavaScript and TypeScript, and secrets scanning supporting both generic secrets and custom patterns, which finds leaked secrets with low false-positive rates.