Features
Built for examinations people cannot afford to lose
Everything below is a property of the engine, not a roadmap item. If it is described here, it is implemented.
Integrity during the examination
- Server-authoritative timing
- Remaining time is computed on the server. Refreshing, reconnecting or changing the device clock cannot alter it, and cannot grant extra.
- A frozen paper
- The question set and its marking rules are snapshotted when an attempt starts. Editing a question afterwards cannot change what a candidate is being marked against.
- Deterministic shuffling
- Where option order is randomised, it is derived from the attempt itself — so a refresh re-presents the identical paper rather than reshuffling it.
- No answer data in flight
- Correct answers and explanations are never sent to the browser while an attempt is open, whatever the client requests.
Resilience
- Autosave on every response
- Answers are written as they are made, each with an operation id, so a retried request can never apply an answer twice or resurrect a stale one.
- Offline queueing
- If the connection drops, responses queue locally and reconcile on reconnect. Replaying the whole buffer is safe.
- Resume where you left off
- Closing a tab mid-paper and returning restores the attempt, its answers and its remaining time.
Marking
- Integer-precise arithmetic
- Marks are held as integer thousandths. No floating-point value ever decides an outcome, so a total is exactly reproducible.
- Partial and negative marking
- Fixed, fractional and percentage penalties; partial credit strategies per question type; sectional and paper-level rules with a defined precedence.
- Objection handling
- A question can be dropped or its key corrected after the fact, with every affected result recomputed and the decision recorded.
- Cutoffs and merit lists
- Category-wise cutoffs, tie-breaking rules and rank generation over a scored cohort.
Learning
- Courses with lessons and quizzes
- Written lessons, attached files and practice quizzes marked by the same engine that marks the examinations.
- Explanations after submission
- A quiz reveals the correct answer and a worked explanation once submitted — never while an attempt is open.
- Progress and certificates
- Per-lesson completion, course progress, and a verifiable completion certificate with a public verification page.
- Reviews and Q&A
- Enrolled learners can review a course and ask questions; instructors answer, and an answer can be marked accepted.
Access
- WCAG 2.1 AA target
- Keyboard operable throughout, no state signalled by colour alone, and the paper magnifies to 5× without breaking its layout.
- No drag-only interaction
- Every question type that could use dragging also has a keyboard path — buttons and position fields, not a drag target alone.
- Bilingual delivery
- Where a paper is offered in English and Hindi, the language can be switched at any point, including for one question.
- Accessibility provisions
- Compensatory time, scribe permission and exemptions are configured per candidate or per category and applied automatically.
What is not built yet
Descriptive question types — essays, précis, letter writing and the rest — are declared in the engine but deliberately not deliverable: they need a human evaluation workflow, and a half-built one would mark a candidate's answer without a marker having read it. The engine refuses to deliver them rather than silently scoring them zero.
Video lessons are served through an external provider behind a swappable interface. An unlisted URL is obscure rather than protected, and instructors are told so.