copilot

GitHub Copilot Extensions is now in Limited Public Beta. With extensions, you can now extend the capabilities of GitHub Copilot Chat and enhance the experience to perform a wide range of actions across third-party tools, services, and data. Create feature flags, check log errors, access API documentation, and even deploy your application to the cloud all through natural language.

We’re starting with GitHub Copilot Extensions from DataStax, Docker, Lambda Test, LaunchDarkly, McKinsey & Company, Octopus Deploy, Pangea, Pinecone, Product Science, ReadMe, Sentry, and Teams Toolkit on the GitHub Marketplace. In the coming weeks, all users will be able to access extensions from Stripe, MongoDB, and Microsoft (including Teams Toolkit and Office) on Visual Studio Marketplace for VS Code as well.

Currently, access to Copilot Extensions is limited to a small set of users and we are planning to gradually roll out larger-scale access as we learn more & gather feedback. To learn more about Copilot Extensions and how to use them, check out the documentation and the GitHub Marketplace.

If you’re looking to build a GitHub Copilot Extension, there are several ways. Join the Copilot Partner Program to explore opportunities to bring your developer tools and services into the GitHub Copilot ecosystem and have extensions light up across all supported clients (VS Code, Visual Studio, and Github.com). If you’re looking to build a VS Code extension specifically, documentation is available here.

Questions or suggestions? Join the conversation in the community discussion.

We’ve updated how we calculate Last Activity to give you better clarity and are pausing access to the Team endpoint in the Metrics API.

Updating the Last Activity calculation

Ahead of the GitHub Copilot Metrics API launch, we made an update to how we calculate Last Activity in order to provide more useful information for admins. Previously, this data point indicated the last time a user generated a Copilot authentication token, which happened automatically if the user’s editor was active. This did not mean the user was engaging with Copilot but rather, the extension was ensuring it could provide completions or chat access if needed.

To align this data point with actual usage, we updated our system to grab the most recent instance where the user deliberately engaged with the Copilot system. These actions include but are not limited to:
– Code completion suggestions show
– Chatting with Copilot Chat in IDEs
– Creating or updating a knowledge base
– Creating a pull request summary
– Interacting with Copilot on GitHub.com

As part of this update, we also needed to perform system cleanup on the vast amount of previous token generated events that were no longer relevant to providing this data point. Some data was erroneously removed but has since been restored.

The Last Activity date should be consistent across the CSV generated via Get Report in Copilot Access settings as well as through the Seats Management API.

Pausing access to Team slicing in the GitHub Copilot Metrics API.

Based on trends in feedback, the product team has learned that the Teams route of the Metrics API returns data that is not meeting the goals of our intended customer experience. In light, they have made the decision to temporarily pull the Teams route from production as of May 9th, 2024. During this time, the team will implement a collection of fixes intended to improve end users’ data experience and plan to re-enable the route by the end of June, at latest.

We understand this may be disappointing but the team is working to restore access as soon as possible. You can provide feedback and follow along for updates via this Discussion.

⏫ Copilot Code Completion model updated with more improvements

We’re excited to announce a new update to the model powering Copilot Code Completion across all IDEs! This update includes improved instruction following and performance improvement for our users. Here are the details:

• Improved instruction following: Copilot can better understand and follow instructions given by the user. This means that Copilot is now better at generating code that matches the user’s intent and requirements.
• Performance improvement: Finally, this model update includes a performance improvement for Copilot users. While this may not be noticeable in all cases, it can help make Copilot even faster and more efficient for certain tasks.

We have introduced two new powerful features to the latest Visual Studio Copilot Chat extension: slash commands (/) that allow you to direct Copilot to perform specific tasks, and context Variables (#) that allow you to specify a file for Copilot to focus its answer on. These features are available on the latest Chat extension that is compatible with Visual Studio 2022 version 17.8 and higher.

Slash Commands

Slash commands are special commands that you can use in chat to receive targeted assistance, including explanations, documentation, test creation, and various other forms of support related to your code. For example, you can use:

• /doc to add a documentation comment
• /explain to explain the code
• /fix to propose a fix for the problems in the selected code
• /generate to generate code to answer your question
• /help to get help with Copilot Chat
• /optimize to analyze and improve the running time of the selected code
• /tests to create unit tests for the selected code

Context Variables

The context variables feature enables you to add files from your solution into your questions using the # symbol. By referencing a file in this way, you allow Copilot to access the content of the file and provide more targeted answers about it. For example, you can ask “How does the #file:’Main.cs’ file work?” or “What is the purpose of the #file:’Calculator.cs’ file?” and get relevant answers from Copilot Chat. You can add multiple files to one question.

Besides these main features, you can also explore our other exciting new preview features

Join the discussion within GitHub Community.

🌐 Upcoming deprecation of Copilot Chat API endpoints

Note: If you are using the latest version of the Copilot extension for Visual Studio or VS Code or you are using the Copilot plugin for JetBrains IDEs, you will not be impacted by this change.

As we announced in September, we updated the API service endpoints used by Copilot Chat. On February 1, 2024, we will deprecate the Copilot Chat API endpoints currently being routed through https://copilot-proxy.githubusercontent.com. Instead, these requests will go through https://api.githubcopilot.com. Versions 0.8.0 and later of the Copilot Chat extension for VS Code and versions 0.1.1817.27579 and later of the Copilot Chat extension for Visual Studio already route chat traffic through https://api.githubcopilot.com. All versions of the Copilot plugin for JetBrains IDEs already route chat traffic through https://api.githubcopilot.com.

To ensure Copilot Chat continues working from February 1, 2024, please update to the most recent version of the Copilot extension and ensure your firewall and network settings allow communication to https://api.githubcopilot.com.

⏫ Update on the Copilot Enterprise Waitlist

On January 19, 2024, we will close the Copilot Enterprise waitlist in anticipation of the general availability release of Copilot Enterprise.

Join the discussion within GitHub Community.

Following our previous communication dated November 8, 2023, regarding the temporary rollback of the Copilot content exclusions feature, we are pleased to announce the re-deployment of this feature with significant enhancements. The rollout will be progressive during the next 10 days as we monitor the behaviour.

With Content Exclusion, GitHub Copilot Business customers will be able to prevent specified files or repositories from being used to inform code completion suggestions made by GitHub Copilot. GitHub Copilot will not be available in excluded files. Organization administrators or repository owners can choose which files or repositories are excluded. Learn more.

Overview of the Issue

Our team observed a critical issue where clients were incorrectly blocked from using Copilot due to the initial implementation of content exclusions. This was primarily caused by errors in fetching content exclusion policies from the client, leading to a temporary suspension of the feature.

Actions Undertaken

In response to this, our engineering team undertook a comprehensive review and rectification process. The issues identified in the client's code were addressed, and additional verifications were implemented on both server and client sides to prevent recurrence.

New Enhancements in the Re-deployed Feature

• Performance Update: We have optimized the performance of the content exclusions feature, ensuring minimal impact on the user experience.
• Extended Coverage: The feature now supports all our official Integrated Development Environments: Visual Studio, JetBrains IDEs, Visual Studio Code, and Vim/Neovim.

Current Status

• Users with pre-existing content exclusion configurations will experience no change.
• New and returning users can now utilize the enhanced feature across all supported IDEs.

Next Steps

We are closely monitoring the performance and user feedback post-deployment. The support for Copilot Chat is also in progress and will be part of the General Availability.

Join the Discussion

We value your feedback and encourage you to participate in the discussion within the GitHub Community.

This month, we made some big improvements to GitHub Copilot! Copilot Chat is now powered by GPT-4 and we updated the model used to detect off-topic chat queries. In VS Code, we are announcing the public beta of code referencing. We also introduced “agents” and the ability to generate commit messages with Copilot. In addition, we improved the context for explaining code and updated the Copilot menu UI. In JetBrains IDEs, we introduced partial acceptance of code suggestions.

Copilot Chat is powered by GPT–4

We upgraded the Copilot Chat experience, bringing more accurate and useful code suggestions with OpenAI‘s GPT–4 model.

Off–topic model improvements for Copilot Chat

As part of our safety features, we‘ve improved our off-topic model to detect chat queries which do not relate to programming. This should result in significantly fewer filtered responses.

Code referencing in VS Code is now in Public Beta

In August, we announced the Private Beta of code referencing in VS Code. This feature searches across billions of files on public GitHub repositories for code that matches a Copilot suggestion. Since then, we’ve heard your feedback, and we’re shipping with a new and redesigned experience. One of the top points of feedback was that the original flow resulted in too many notifications. To fix this, if there’s a match, users will find its information displayed in the Copilot console log, including where the match occurred, any applicable licenses, and a deep link to learn more. If you are interested in code references, you can refer to the window, otherwise, it won’t be in your way.

The deep link will now take you to a navigable page on GitHub.com to browse examples of the code match and their repository licenses, and see how many repositories — including ones without licenses — that code appears in, as well as links to those repositories.

Learn more about Copilot code referencing and let us know your thoughts in the GitHub Community!

Introducing “agents” in Copilot Chat in VS Code

We have introduced a new capability called “agents” to enhance your interaction with Copilot Chat. Agents are like specialized experts who can assist you with specific tasks. You can mention them in the chat using the @ symbol. Currently, there are two agents available:

• @workspace: This agent has knowledge about the code in your workspace and can help you navigate it by finding relevant files or classes. The @workspace agent uses a meta prompt to determine what information to collect from the workspace to help answer your question.
• @vscode: This agent is knowledgeable about commands and features in the VS Code editor itself, and can assist you in using them.

Each agent also supports slash commands. The slash commands you may have used before should now be used with an agent. For example, /explain is now @workspace /explain.

Read more in the VS Code release notes.

Improved explanation context in Copilot Chat in VS Code

You can ask Copilot Chat to explain a code selection in your active editor either through the @workspace /explain command or through the “Explain with Copilot” action in the context menu. Copilot Chat has now integrated implementations of referenced symbols, such as functions and classes, which leads to explanations that are more precise and useful. This works best across files when you have an extension contributing language services installed for one of the following languages: TypeScript/JavaScript, Python, Java, C#, C++, Go, or Ruby.

Commit message generation using Copilot in VS Code

Copilot can now generate commit messages based on the pending changes using the new “sparkle” action in the Source Control input box.

Updated Copilot menu in VS Code

Our Copilot menu in VS Code is now more visible and aligned with our design for JetBrains IDEs. It is now easier to understand the current status of Copilot, access the various settings or documentation.

The new menu is displayed when clicking on the Copilot icon in the lower right corner in the statusbar of VS Code.

JetBrains partial acceptance for code suggestions

The Copilot extension for JetBrains IDEs has leveled up! You now have the flexibility to incorporate code suggestions piece by piece, whether that’s word-by-word or line-by-line. Feel free to customize these shortcuts as you prefer. Happy coding!

We welcome your feedback on Copilot! Please join the discussion in the GitHub Community.