How Copilot Spaces gives your AI the right project context | GitHub Checkout

Key Concepts

• GitHub Copilot Spaces: A feature that provides developers with a contextualized environment for working with GitHub Copilot, allowing it to access project-specific information.
• Contextualization: The process of providing relevant information (files, pull requests, issues, repos) to Copilot to improve the accuracy and relevance of its responses.
• CodeQL: A popular open-source project at GitHub used as a case study for demonstrating Copilot Spaces.
• "check_call" vulnerability: A specific security issue within the CodeQL codebase that serves as the example problem.
• Copilot Coding Agent: An agent within Copilot Spaces that can propose and generate code changes.
• GitHub MCP Server: A server that enables access to GitHub Copilot Spaces directly within an Integrated Development Environment (IDE).
• Public API: A planned feature for Spaces that will allow programmatic access and integration.
• Use Cases: Code generation, planning issues/projects, and knowledge sharing.

GitHub Copilot Spaces: A Contextualized Development Environment

GitHub Copilot Spaces is presented as a novel approach to enhance developer workflows by providing GitHub Copilot with rich, project-specific context. This aims to overcome the common developer pain point of time-consuming manual information gathering before addressing issues or tasks.

Core Functionality and Contextualization

Spaces allow developers to bring together various sources of information that Copilot needs to generate relevant and grounded responses. These sources include:

• Files: Specific files or entire repositories can be added to a Space.
• Pull Requests (PRs): Linking PRs provides Copilot with information about proposed changes and discussions.
• Issues: Issues can be directly linked, allowing Copilot to understand the problem being addressed.
• Repos: The entire repository can be included for comprehensive context.

A key feature is the dynamic updating of linked sources. Once a URL is pasted into a Space, the content remains current as the underlying files or issues on GitHub are updated, eliminating the need for manual refreshes.

Content Inclusions and Data Types

Spaces can consume a wide array of information:

• Entire Repositories: For broad context.
• Specific Files: Curated selection of relevant files.
• Linked Files, PRs, and Issues: Via URL pasting.
• Uploaded Text Content: This includes notes, Slack messages, and even video transcripts.
• Future Support: Planned support for .doc/.docx and .pdf file types.

Demonstrating Spaces with a Real-World Example: CodeQL

The transcript details a practical demonstration using the CodeQL open-source repository.

Scenario: A maintainer receives an issue reporting an unsafe usage of "check_call" within the codebase. The maintainer is unsure of the best solution.

Steps using Spaces:

1. Space Creation: A dedicated Space is created for the CodeQL project.
2. Adding Sources:
   • Design patterns.
   • Security guidelines.
   • Accessibility recommendations.
   • The entire CodeQL repository.
   • The specific issue URL is pasted, which is then automatically listed.
3. Defining Instructions for Copilot:
   • Persona: Act as an experienced engineer.
   • Grounding: Ensure responses are based on the Space's content.
   • Output Format: Explain the goal, approach, and execution, and create a plan before acting.
   • Triggering Agent: Instruct Copilot to trigger its coding agent after planning.
4. Interacting with Copilot:
   • The prompt "Help me debug this issue" is given to Copilot.
   • Copilot analyzes the linked issue and the Space's content.
   • Copilot outlines its plan: sanitize inputs and disable SSH.
5. Proposing Code Changes:
   • The user requests "Propose code changes using Copilot coding agent."
   • Copilot confirms the action.
   • Upon approval, the Copilot Coding Agent generates a pull request.
6. Pull Request Generation:
   • The generated PR includes "before" and "after" code views.
   • Crucially, the PR references the specific sources within the Space from which the information was derived, allowing for validation.

Iteration and Collaboration

• Iterative Development: Developers can iterate on Copilot's suggestions within the Space by providing further instructions, leading to new PRs with updated code.
• Knowledge Collaboration: Spaces are designed for collaborative knowledge sharing.
  • Privacy: Spaces are private by default.
  • Sharing Options: Users can share Spaces with their entire organization, specific individuals, or control access through enterprise admin settings, ensuring security.

IDE Integration and Future Roadmap

• IDE Access: GitHub Copilot Spaces are now accessible within IDEs via the GitHub MCP Server. Users need to download the MCP Server to access the Spaces toolset.
• Roadmap:
  • Public API: Planned for release, enabling programmatic integration.
  • Additional File Types: Support for image files, .doc/.docx, and .pdf is being explored.

Key Use Cases for Teams

The transcript highlights three primary use cases for GitHub Copilot Spaces:

1. Code Generation: As demonstrated with the Copilot Coding Agent for debugging and feature implementation.
2. Planning Issues and Creating Projects: Product managers can use Spaces to brainstorm features, detail requirements, and link relevant repos, issues, and features. Engineers can then leverage these same Spaces for code generation.
3. Knowledge Sharing: Teams can create Spaces to document how their codebase works, including code files, issues, and even video transcripts of architecture walkthroughs. This allows anyone within the company to query Copilot Chat for information about the codebase or specific features.

Conclusion

GitHub Copilot Spaces aims to revolutionize developer productivity by providing a contextualized and collaborative environment. By integrating project-specific information directly with Copilot, it streamlines debugging, planning, and knowledge sharing, ultimately fostering more efficient and informed development workflows. The ongoing development, including IDE integration and expanded file type support, signifies a commitment to making Spaces a central tool for modern software development teams.