SafeSight
AI that teaches, not answers. SafeSight gives every student a tutor that builds real understanding, and gives every teacher a clear view of how AI is used in their class. Built at IIT Mandi.
Status: core build done (S1–S12: Sage, Lens, Trace, teacher dashboard, landing site, demo data). See SESSIONS.md for what's next and docs/DEMO.md for the demo runbook.
What's in this repo
| Path | Purpose |
|---|
PLAN.md | Product and build plan: scope, agents, design system, architecture, database, privacy, session prompts, demo script |
CLAUDE.md | Working rules and running status for Claude Code sessions |
SESSIONS.md | Build tracker: one Claude Code session per feature, S1 → S15 |
design/ | Brand book, tokens, logo and the finished landing-page prototype |
.env.example | Every environment variable the app will need |
Stack
Next.js (App Router, TypeScript) · Tailwind + shadcn/ui · Motion · Recharts · Supabase (Postgres, Auth, Storage, pgvector) · Vercel AI SDK with an OpenAI-compatible model provider · Tiptap · zod · Vercel hosting.
Working on it
All building happens in Claude Code cloud sessions on this repo:
- Open Claude Code on the web, pick this repo.
- Start with:
Read CLAUDE.md and PLAN.md first. then paste the next prompt from SESSIONS.md.
- Review the PR and the Vercel preview on phone and laptop, then merge.
Local development (needs Node 20+, pnpm and Docker):
pnpm install
cp .env.example .env.local # fill in values (local keys come from `pnpm db:start`)
pnpm db:start # local Supabase in Docker, applies supabase/migrations
pnpm seed # demo institution, teacher, 2 courses, 30 students, admin
pnpm dev
Checks: pnpm lint, pnpm typecheck, pnpm test, pnpm build.
Team
SafeSight is a student startup from the Essentials of Entrepreneurship course at IIT Mandi.
© 2026 SafeSight. All rights reserved.
