- Grader now passes all rounds of user answers to LLM (tagged 第N轮回答)
- LLM can see what was already answered and avoid redundant follow-ups
- Updated all three language prompts with multi-round guidance
- Option display: use slice(1) instead of regex to strip letter prefix
- Grader prompts: add explicit partial credit guidance (5-7 for partial, 0-2 only for off-target)
- Grader: separate followup hint from scoring feedback
- Interviewer: use followup hint directly without prefix/suffix
- Restored standard and choice question presentation paths
- selectQuestions: shuffle final result for random question order
- grader: wrap LLM invoke in try-catch, default score 5 on failure
- grader: inner try-catch for JSON parse errors, graceful fallback
- Data pathway: flow options through questions, answerKey in graph state
- Interviewer: format MULTIPLE_CHOICE with A/B/C/D options
- Grader: instant choice scoring (zero LLM), compare correctAnswer
- AssessmentView: render choice buttons vs textarea based on questionType
- Security: sanitizeStateForClient strips correctAnswer/judgment/answerKey
- Bank detection: check PUBLISHED items (not PUBLISHED bank status)
- Batch UI: select all / batch approve / batch reject on detail view
P0-1 Backend: dimensions column on template entity + validation
P0-1 Frontend: dimensions edit UI in TemplateManager
P0-2: routeAfterGrading unit tests (10 cases), service spec fix + certificate tests, jest-e2e.json
P1-1: proper PDF generation with embedded CJK font via pdf-lib low-level API
- Add pagination support to findAll (page, limit query params)
- Add findByTemplateId method to service
- Add GET /by-template/:templateId endpoint to controller
- Service already includes CRUD for QuestionBank and QuestionBankItem
Developer