🌊 Nalu

Learn anything. AI-powered. Structured.

Nalu is an AI-powered learning platform that generates personalised, conversational courses on any topic. Pick a subject, get assessed, and learn through dialogue with an adaptive tutor that tracks your progress, schedules spaced review, and awards XP as you grow.

Think Duolingo, but for anything.

How It Works

1. Choose a topic — "Python programming", "South African wines", "music theory", anything.
2. Scope it — Nalu asks a few clarifying questions to tailor the curriculum.
3. Get assessed — A baseline quiz determines what you already know.
4. Learn through conversation — Nalu teaches, asks questions, surfaces blindspots, and adapts to your direction.
5. Build knowledge — Spaced repetition brings back concepts at the right time. XP tracks your progress across proficiency tiers.

Tech Stack

• Framework: Next.js 16.2 (App Router, Turbopack)
• Language: TypeScript (strict mode)
• API: tRPC v11
• Database: Supabase (PostgreSQL)
• LLM: OpenAI-compatible API (provider-swappable)
• Styling: Tailwind CSS
• Testing: Vitest + Playwright
• Task runner: just
• Package manager: Bun

Getting Started

Prerequisites

• Bun >= 1.0
• just (brew install just)
• Supabase account (free tier)
• Cerebras API key (free tier) or any OpenAI-compatible provider

Setup

git clone <repo-url>
cd nalu
bun install

# Environment
cp .env.local.example .env.local
# Fill in your Supabase and LLM provider credentials

Run

just dev        # Start dev server (localhost:3000)

Dev mode (NEXT_PUBLIC_DEV_MODE=true in .env.local) enables auto-login and skips auth for rapid testing.

Development

just dev        # Start dev server
just test       # Run unit tests
just test-watch # Run tests in watch mode
just lint       # Lint
just typecheck  # Type check
just check      # Run all checks (lint + typecheck + test)
just build      # Production build

Project Structure

src/
  app/              # Next.js pages and routes
  server/routers/   # tRPC API layer
  lib/
    llm/            # LLM client (provider-agnostic)
    prompts/        # All prompt templates (single source of truth)
    spaced-repetition/  # SM-2 algorithm and review scheduling
    scoring/        # XP calculation and tier advancement
    course/         # Course state management
    types/          # Shared TypeScript types
  components/       # React UI (thin rendering layer)
  db/queries/       # Typed database access
tests/
  e2e/              # Playwright end-to-end tests
  fixtures/         # Mock LLM responses
docs/
  PRD.md            # Full product requirements document

Architecture rule: all business logic lives in src/lib/. Components render. Routers orchestrate. See CLAUDE.md for the full set of conventions.

Design

Visual identity inspired by Hokusai's Great Wave off Kanagawa. Glassmorphic UI with a clean, modern feel. Learning blocks are called "waves." The XP progress bar is wave-shaped.

Licence

Contributing

This is currently a solo project in early MVP. If you're interested in contributing, open an issue to discuss.