VS Code Extension for Children
| ID | Requirement | Priority | Completed | Contributors |
|---|---|---|---|---|
| 1. | Provide code templates implementing basic game mechanics. | Must |
|
Zoltan |
| 2. | Offer explanations for methods, classes, and code sections | Must |
|
Zoltan |
| 3. | Automatically detect errors and suggest corrections in real-time | Must |
|
All |
| 4. | Image Asset Generation | Must |
|
Jash |
| 5. | Walkthrough Implementation | Must |
|
Alicia |
| 6. | Full support for MakeCode and Unity for both 2D and 3D game development | Must |
|
Maulik |
| 7. | Generated content must be suitable and safe for users under 18, with explicit moderation to exclude inappropriate content | Must |
|
All |
| 8. | Advanced Real-Time Suggestions (extended) | Should |
|
Jash |
| 9. | Clear, structured instructions tailored to beginner skill levels | Should |
|
All |
| 10. | Proactive clarifying questions to validate AI-generated outputs | Should |
|
Zoltan |
| 11. | Ability to interact with and test gameplay live | Could |
|
N/A |
| 12. | AI-assisted generation of special effects within games | Could |
|
N/A |
| 13. | Advanced multiplayer game logic or extensive backend multiplayer support | Won't |
|
N/A |
| ID | Bug Description | Priority |
|---|---|---|
| 1. | GitHub oAuth does not work. | High |
| 2. | Unexpected outputs from the AI Models for different prompts. | Medium |
| 3. | When using ‘@pixel /findVideos’ for the first time when the extension is reloaded, it gives an error message that no text is selected even if the text is selected correctly. However, without changing the selection or touching anything else, if we re-run the same prompt, it works correctly. | Low |
| 4. | While using the Qwen model, it outputs the base prompt in the chat window before giving the actual output | Medium |
| Work Packages | Zoltan | Maulik | Jash | Alicia |
|---|---|---|---|---|
| Project Partner Liaison | 100.00% | 0.00% | 0.00% | 0.00% |
| Requirement Analysis | 65.00% | 10.00% | 25.00% | 0.00% |
| HCI | 0.00% | 20.00% | 40.00% | 40.00% |
| Research & Experiments | 30.00% | 30.00% | 30.00% | 10.00% |
| UI Design | 0.00% | 0.00% | 0.00% | 100.00% |
| Coding | 33.00% | 33.00% | 34.00% | 0.00% |
| Testing | 28.80% | 40.00% | 31.20% | 0.00% |
| Project Website | 0.00% | 20.00% | 40.00% | 40.00% |
| Presentation Planning | 0.00% | 30.5.00% | 36.5.00% | 33.00% |
| Video Editing | 0.00% | 50.00% | 20.00% | 30.00% |
| Overall | 25.68% | 23.35% | 25.67% | 25.3% |
We aimed to create a friendly and approachable user interface that would appeal to children and beginner developers. Key considerations included simplicity, vibrant visuals, and intuitive navigation for the Walkthrough. The layout across all tabs — Home, History, Voice-to-text and File Attachment — was kept consistent to avoid confusion and reduce the learning curve.
Feedback from user testing sessions indicated that the interface was easy to understand and fun to use. However, there is room for improvement in terms of accessibility (e.g. keyboard navigation and screen reader support) and making the interface more responsive across different screen sizes.
We give our UI/UX a rating of Good.
PixelPilot’s core functionality — generating and managing code snippets via GitHub Copilot integration, generating images and real-time suggestions — worked reliably during testing. All major features were functional, and no critical bugs were reported during demonstrations or internal tests.
While certain advanced suggestions or contextual understanding by the Copilot could be improved, this is largely dependent on the underlying AI model rather than our implementation.
We rate our functionality as Good.
Our project was built with maintainability in mind. By modularizing features into reusable components and maintaining a consistent file structure, we ensured that future changes could be made with ease.
The codebase includes comments and documentation to support onboarding new developers. Our use of TypeScript interfaces and utility functions further contributed to cleaner, more understandable code. A README file and developer setup instructions were also provided.
We rated our maintainability as Good.
To ensure stability, we conducted both unit testing and manual user acceptance testing. Unit tests covered key components like code visualization and input handling, while manual tests helped us uncover minor UI issues across various environments.
The application ran consistently across desktop environments, though some issues with resizing and minor layout glitches were noted. More time would have allowed for more extensive automated testing coverage.
We rated our stability as Adequate.
Since PixelPilot is designed to be lightweight and locally operated, the performance is generally fast and responsive. Image generations appear with minimal delay, and the code suggestions respond promptly to changes.
There are no heavy assets or external API calls beyond GitHub Copilot, which helps reduce loading times. On slower internet connections, some delay was noticed when fetching Copilot suggestions.
We rated our efficiency as Good.
PixelPilot works best on desktop devices and has been tested on Chrome, Firefox, and Edge. Since the tool is geared toward code development, it is not designed with mobile or tablet usage in mind.
That said, the app performed well on various resolutions within desktop browsers. Future work could focus on improving compatibility with tablets or touch-screen interfaces for broader accessibility.
We rated our compatibility as Adequate.
We maintained steady progress through effective use of GitHub Projects and weekly team syncs. Tasks were divided fairly across team members, and we practiced pair programming during complex feature development.
GitHub Issues and Pull Requests helped us track bugs, manage features, and review each other’s code. While we initially planned to use a Trello board, we found GitHub’s built-in tools more suited to our workflow.
We rate our project management as Very Good.
If we had more time we would improve/extend the project in the following ways:
One of our main goals moving forward would be to enhance how the Copilot integration handles contextual information. Currently, prompts are generated based on individual code snippets, but there's potential to implement a smarter context engine that tracks files, game state, and user history more effectively. This would result in more accurate and useful suggestions, especially for longer or more complex projects that children might attempt.
Given our target audience includes children and beginners, the next feature we would like to develop is an interactive tutorial mode. This mode would guide users through simple steps in game development, such as creating a sprite or adding movement. Tutorials could be gamified and include visual cues, making the learning process more engaging and intuitive.
Currently, all files and history exist only within the session. A future improvement would be to implement a user project system where users can save, load, and organize their projects. This would enable persistence across sessions and encourage users to return and iterate on their work over time.
Long-term, we envision integrating a visual builder where users can drag and drop elements like characters, objects, and logic blocks into their game world. This would reduce the reliance on typing and make the platform more inclusive for younger children or users unfamiliar with code syntax.
Discover additional resources, documentation, and supplementary materials for Pixel Pilot.