What is GitHub Copilot Spaces? Centralize your project’s context | GitHub Checkout

GitHub Copilot Spaces: A Deep Dive

Key Concepts:

• Copilot Spaces: A new feature for organizing, sharing, and accessing institutional knowledge within a company using contextualized AI.
• Contextual AI: AI responses grounded in a specific, curated set of information, reducing hallucinations and improving accuracy.
• Inner Sourcing: Sharing internal knowledge and resources across teams within an organization.
• Subject Matter Expertise (SME) Multiplication: Leveraging AI to scale the knowledge and best practices of experts across an organization.
• Accessibility Compliance: Ensuring software and systems meet accessibility standards for users with disabilities.
• IDE Integration: Bringing Copilot Spaces functionality directly into Integrated Development Environments for real-time assistance.

What are Copilot Spaces?

Copilot Spaces provide a novel way to interact with codebases and company-specific context. They facilitate the organization and sharing of information related to code, documentation, or even abstract knowledge like transcripts from deep-dive sessions. The core functionality involves creating a "Space" containing various contextual elements, enabling users to chat with the AI specifically within that context. This approach yields high-quality results due to the focused context management. Spaces can be shared with teammates, fostering an inner-sourcing model for company knowledge.

Content and Configuration

Spaces can be customized with:

• Name and Description: To clearly identify the purpose and scope of the Space for other users.
• Custom Instructions: Providing context to Copilot about the Space's purpose and desired response style.
• Attachments:
  • Files from Repositories: Code files, HTML, documentation, and other relevant files stored in GitHub repositories.
  • Text File Content: Allows adding any text-based information, including Slack threads, video transcripts, and code snippets. This is highlighted as a creative workaround for incorporating diverse information sources.
  • Issues and Pull Requests: (Coming Soon) Integration with GitHub issues and pull requests to provide context around specific development tasks.

Use Cases and Examples

• Accessibility Space (GitHub Example): Katherine, an accessibility SME at GitHub, created a Space containing various accessibility documents. This Space allows developers to ask questions about accessibility compliance (e.g., "What is mas c compliance?") and receive high-quality, grounded answers, effectively making Copilot an accessibility expert. The Space link is pinned in Slack and included in accessibility review workflows.
• Datadog Logging Space: Defining a Space with examples of how to implement Datadog logging, including examples of good logs, to help engineers scaffold out logging for new features.
• Meta Space (Spaces about Spaces): A Space containing the code for Copilot Spaces, along with video transcripts from deep-dive sessions with engineers. This allows GitHub employees to ask questions about the architecture, roadmap, and inner workings of Copilot Spaces, receiving answers grounded in both code and design discussions.

Benefits and Advantages

• High-Quality Results: By grounding responses in a specific context, Copilot Spaces minimize hallucinations and provide more accurate and relevant answers.
• Security and Privacy: The contextualized approach ensures that information remains within the defined Space, addressing security concerns about sensitive or proprietary information being exposed.
• Knowledge Sharing and Inner Sourcing: Spaces facilitate the sharing of expertise and best practices across teams, promoting a more collaborative and efficient development environment.
• Onboarding and Training: Spaces can be used to onboard new team members by providing them with access to curated knowledge and resources.
• Subject Matter Expertise Multiplication: Allows SMEs to scale their knowledge and impact by creating Spaces that capture their expertise and make it accessible to others.

Access and Entitlements

• Availability: Copilot Spaces are available to anyone with a Copilot license, including Copilot Free users and Enterprise customers.
• Entitlement Model: Using Spaces with the base model does not count against entitlement usage. However, using them with a premium model may consume entitlements, depending on the specific model.

Future Development

• IDE Support: Integration with IDEs is planned, allowing developers to access and utilize Spaces directly within their development environment. This will enable real-time assistance and guidance based on the context of the Space.
• Community Feedback: GitHub is actively soliciting feedback from users through a public discussion post and encourages businesses and enterprises to work with their admins and GitHub contacts to provide input.

Revoking Access

Access to a Space can be easily revoked by changing the access settings back to "No access."

Notable Quotes

• Kelly: "Spaces are the future of subject matter expertise within your company."
• Kelly: (Regarding text file content attachments) "This is really the greatest hack out there because you can put anything into here."

Technical Terms

• Hallucinations: Instances where AI models generate incorrect or nonsensical information.
• Context: The specific set of information that an AI model uses to generate responses.
• MAS Compliance: Meeting the requirements of the Massachusetts Accessibility Standards, a set of guidelines for making websites and applications accessible to people with disabilities.
• IDE: Integrated Development Environment, a software application that provides comprehensive facilities to computer programmers for software development.

Logical Connections

The video logically connects the problem of knowledge silos and difficulty accessing internal expertise with the solution offered by Copilot Spaces. It demonstrates how Spaces address these challenges by providing a centralized, contextualized, and shareable knowledge repository. The examples provided illustrate the practical applications of Spaces in various scenarios, such as accessibility compliance, logging implementation, and internal knowledge sharing. The discussion of future development plans, particularly IDE integration, highlights the ongoing evolution and potential of Copilot Spaces.

Synthesis/Conclusion

GitHub Copilot Spaces represent a significant advancement in how organizations manage and share internal knowledge. By providing a contextualized AI experience, Spaces enable users to access high-quality, relevant information, reduce hallucinations, and promote collaboration. The ability to incorporate diverse content types, including code, documentation, and video transcripts, makes Spaces a versatile tool for capturing and disseminating expertise. With planned IDE integration and ongoing community feedback, Copilot Spaces are poised to become an essential component of the modern software development workflow, empowering teams to build better software more efficiently.