Getting Started with Prompts in the GitHub Copilot App
GitHub has released a guide on writing effective prompts within the GitHub Copilot app, emphasizing model selection and context management. It details the operational changes and compatibility requirements when transitioning to the updated interface and APIs.
The latest updates to the GitHub Copilot app introduce structured workflows for writing prompts, managing input context, and selecting the most appropriate underlying models. Users can now leverage advanced evaluation criteria to assess how different models interpret specific prompts, ensuring more predictable code generation. The update outlines the functional differences in API structures, making it easier to integrate Copilot deeper into existing development workflows.
Primary source
Open GitHub Blog ↗This article summarizes the source. Check the original for full details and conditions.
Comparison
| Aspect | Before / Alternative | After / This |
|---|---|---|
| Model Selection | Implicit or automated model selection with limited user control over the underlying LLM. | Explicit model selection tailored to specific tasks and complexity requirements within the app. |
| Context Management | Implicit workspace tracking without explicit configuration or visibility. | Structured context provision allowing engineers to define precise boundaries and files. |
| API Compatibility | Legacy API endpoints with standard response structures. | Updated endpoints incorporating advanced prompt metadata and evaluation metrics. |
Action Checklist
- Verify API endpoint compatibility with the new GitHub Copilot app specifications Legacy payload structures might require minor mapping updates
- Select the appropriate context boundaries in your workspace configuration Restricting the context size prevents unnecessary token usage and latency
- Test prompt evaluation outputs against your project's coding standards Verify if the selected model matches the security and formatting guidelines of your team
