Approximately 90% of the source code was written by AI, there is no point in being slow and inefficient simply to manually code everything. What matters is delivering value to people.
Mobile-first, open-source knowledge platform for Bangladesh national curriculum students.
- Class → Group → Subject → Chapter → Topic curriculum hierarchy in Supabase
- Topic-wise article system with Markdown + LaTeX support
- Contributor tracking per article
- Reputation-based contribution workflow
- Proposal and approval flow for low-reputation users
- Product suggestion system where any user/visitor can submit feature requests, updates, and bug reports
- AI actions via Groq GPT OSS 120B:
- Generate missing article drafts
- Review and improve existing articles
- Step-by-step problem solver
- Topic recommendation from a student problem description
- Google OAuth-only authentication (via Supabase)
- Profile page (reputation, edit history, proposals)
- Admin panel:
- Monitor users
- Manually set reputation
- Update user roles
- Review proposals
- Review/manage product suggestions (status, priority, notes)
- Manage class/group/subject/chapter/topic records
- Next.js App Router + TypeScript
- Supabase (Postgres + Auth + RLS + RPC)
- Groq API using GPT OSS 120B
react-markdown+remark-math+rehype-katex
- One topic = one article.
- Final model is topic-only: content is stored in
curriculum_topics.body_markdown. - No separate
articlestable is used in the final schema, so duplicate article rows per topic are impossible. - Topic
titleis the content title; topicbody_markdownis the content body.
- Install dependencies:
npm install- Configure environment:
cp .env.example .env.localFill all values in .env.local.
-
In Supabase SQL editor, run files in this order:
-
supabase/migrations/202602200001_initial.sql -
supabase/seed/202602200002_curriculum_seed.sql -
supabase/seed/202602200003_content_seed.sql -
In Supabase Auth settings:
- Enable Google provider
- Set callback URL to:
https://YOUR_DOMAIN/auth/callback(orhttp://localhost:3000/auth/callbackfor local)
- Run dev server:
npm run devNo email/password flows are implemented. Contribution actions require Google OAuth session.
- High-reputation users can publish directly.
- Others can still propose changes.
- Qualified reviewers approve/reject proposals.
- Rewards are tracked in
reputation_eventsand aggregated inuser_profiles.reputation. - Thresholds and reward values are configurable in
reputation_policy.
AI-generated content is intentionally routed through contributor moderation flow to keep factual quality aligned with syllabus expectations.