KUCODE
Summary
To connect coursework with development, I built a platform that links the GitHub accounts of students at Korea University’s College of Informatics with academic records, making course-level GitHub activity visible in a way repositories alone cannot. It lets professors and teaching assistants analyze activity by course, and lets students use their accumulated activity as an e-Profile and portfolio material.
- Role
- Project LeadPMFull-stack Developer
- Outcome
- 1,018 students in analytics9,095 repositories30 linked course offerings
Making GitHub Activity Useful Beyond the Classroom
Students’ code and collaboration histories are recorded on GitHub. But with activity spread across repositories, it was difficult to bring it together by course or summarize an individual student’s projects and contributions. Faculty and teaching assistants needed to review students’ development activity by course, while students needed to use that history in their portfolios.
KUCODE was built to meet these needs by linking the GitHub accounts of students at Korea University’s College of Informatics with academic records. Faculty and teaching assistants can analyze development activity by course, while students can use their own activity in e-Profiles and portfolio material. I led the project and worked on full-stack development with Django and Vue.js, as well as setting up the deployment infrastructure.
| Measure | Count | Scope |
|---|---|---|
| Students in analytics | 1,018 | Deduplicated by student ID |
| Linked repositories | 9,095 | Linked to student accounts; includes forks |
| Linked course offerings | 30 | 5 semesters; includes sections |
A Data Structure Linking 1,000+ Students, 30 Course Offerings, and 9,000+ Repositories
To compare GitHub activity by course, repositories and student accounts needed to be linked with enrollment information. The same course can have different students and projects depending on the year, semester, and section. We designed the data structure to manage courses, enrollment histories, and project repositories separately and link them so that a student’s project participation could be viewed by course.
GitHub IDs link development activity to students, while year, semester, and course code distinguish course offerings. Enrollment records link students to courses, and project records link courses to repositories, allowing activity to be viewed by course separately from each student’s accumulated activity.
Showing a team repository’s total commit count does not distinguish an individual student’s contribution. Commit, PR, and issue authors are therefore stored separately from repository contributor information, and the e-Profile shows that student’s own activity. It also distinguishes the student’s own repositories from contributions to other repositories.
Syncing Only Changed Repositories Instead of Recollecting Everything
The service links 9,095 repositories to student accounts. Recollecting all their activity on each run would repeat API calls and updates for unchanged data. To reduce that work, the current implementation compares repository update and push timestamps to select repositories for collection. It also reuses each owner’s repository listing within a synchronization run to reduce repeated lookups.
However, PRs and issues can be created or change state even without a code push. Selecting repositories solely by their push timestamps could miss this activity, so PR and issue changes are checked separately even when repository metadata appears unchanged. Repositories with activity that needs updating are included in the collection list.
An API failure does not mean that no activity has changed. The repository therefore stays in the collection list to avoid skipping activity because the check failed. A full collection option remains available, and a uniqueness constraint on repository and SHA prevents duplicate commit records.
Course-level GitHub Activity Without Opening Repositories One by One
The course view brings together enrollment and counts of commits, PRs, issues, stars, and repositories so faculty and teaching assistants can review development activity without opening each repository. Academic records and development activity can be viewed together, and the analytics screen lets users switch the scope by course or department.
Roster import, search, and statistics export also support course administration. Course management shows offerings by semester, while student analytics brings together enrolled courses and GitHub activity so users can review a student’s course participation and development activity.
Data collection is separate from page requests so opening a page does not require waiting for GitHub API responses. Analytics APIs read stored activity from the database. In student–course analytics, prefetching enrollment and course records and reusing repositories grouped by GitHub ID and projects grouped by repository ID reduced repeated lookups while processing students and courses.
Turning Collected GitHub Activity into Student e-Profiles and Portfolios
GitHub activity collected for educational administration is also provided through individual e-Profiles so students can use it to describe their own experience. Students can review their main programming languages, commits, PRs, issues, activity trends, and project lists, alongside an introduction and technology stack.
Students can review course projects and independent activity separately and choose which projects to include in a portfolio PDF. They can choose projects relevant to their target field or the experience they want to describe and use them as portfolio material.
KUCODE Owned from Planning Through Development and Operations
As project lead and PM, I organized requirements from the development plan and screen specifications and took on full-stack development and infrastructure setup. Students organizing their own activity and faculty and teaching assistants analyzing courses use the same GitHub data for different purposes. I designed the screens around the information and activity scope each user group needed.
The service was built with Django and Vue.js and uses PostgreSQL as its database. Nginx and Gunicorn were configured through Docker Compose. Development, staging, and production use separate Compose projects and data volumes to keep development configuration and data from being mixed with production.
KUCODE holds data for 30 linked course offerings across 5 semesters. Faculty and teaching assistants can analyze development activity by course using academic records and GitHub activity together, while students can organize their project experience into portfolio material.