Appearance
System Architecture
CS14 is a full-stack research platform: researchers design social-media-style survey studies in the admin workspace, participants complete the study through a public share link, and the backend stores survey content, responses, calibration metadata, gaze samples, click records, and exports.
Runtime View
Main Modules
| Module | Responsibility | Important files |
|---|---|---|
| Admin workspace | Create/edit/publish surveys, configure platform style, groups, posts, translations, preview, analytics, export. | frontend/app/admin/* |
| Participant runner | Consent, language selection, calibration, social feed rendering, click/gaze collection, completion. | frontend/app/survey/[shareCode]/* |
| Backend API | Auth, survey CRUD, public participant flow, translations, analytics, export, tracking ingestion. | backend/app/routers/* |
| Data model | Researcher-owned surveys, posts, questions, translations, anonymous responses, calibration and tracking records. | backend/app/models/* |
| Docs/runtime ops | Public documentation, demo runbook, deployment checklist, acceptance matrix. | docs-site/docs/*, docs/* |
Participant Data Flow
Design Choices
- Public participant links stay anonymous. The participant token binds updates to one response without requiring login.
- Researcher operations require auth and ownership checks. Analytics, exports, translations, and survey editing are not public.
- Platform style is a research variable.
platform_ui_stylecontrols the participant visual treatment whileplatform_styleremains a legacy compatibility field. - Camera data is minimized. The browser uses webcam frames locally for MediaPipe detection, but the backend stores numeric calibration/gaze records and confidence summaries, not raw video.
- Exports are research-first. CSV/JSON include response metadata, group/language filters, calibration status, attention confidence, interactions, gaze, and click evidence.