Nirdeshika
Gamified AI PWA for studying Nepal Rastra Bank directives: quizzes, streaks and a community feed.
Nirdeshika is a study companion for people in Nepal's banking sector who have to know Nepal Rastra Bank directives and circulars inside out.
Problem
NRB directives are long, change often, and are usually studied by re-reading PDFs. There is no easy way to check what you actually remember, see what changed in a new circular, or keep a daily study habit going.
What I built
A gamified, installable web app. Admins upload a public NRB circular and Gemini turns it into a quiz with an explanation for every answer. Learners practise in quiz mode, build daily streaks, earn XP, compete in a weekly league, and discuss circulars in a community feed. It installs to the home screen as a PWA, and sign-in supports passkeys as well as a password.
Feature reel
Streaks & leaderboard
Daily streaks and a weekly league keep bankers studying NRB directives every day.
Community feed
Ask questions, share notes and discuss circulars with other learners.
Login + install
Sign in once, then install it to the home screen like a native app.
Architecture & stack
- Next.js 16 (App Router) with server-rendered pages and route handlers
- PostgreSQL through Prisma, hosted on Neon
- better-auth for email/password sign-in, usernames and passkeys; sign-up is closed and accounts are created by an admin
- Gemini for quiz generation and per-question explanations; if the model is unavailable, the document is still stored and can be processed later
- A hand-written service worker and web manifest for the PWA, so the core pages work like an app
What I learned
AI generation has to be treated as an unreliable dependency: the upload path keeps working when the model is overloaded instead of failing the whole request. Gamification works best when it is built on activity the app already records, so streaks, XP and the weekly league come from quiz attempts rather than separate tracking.