BiteBase

An interactive micro-learning tutor that swaps doomscrolling for bite-sized growth. Tell it what to learn, and it builds a playful, quiz-filled course tailored just for you.

Stack

Layer	Tech
Monorepo	Turborepo + pnpm workspaces
Web	Next.js 15 (App Router)
Mobile	Expo SDK 56 (React Native)
Styling	Tailwind CSS + NativeWind
API	tRPC v11
Database	PostgreSQL + Drizzle ORM
Auth	Better Auth
AI	Vercel AI SDK → Ollama (local, OpenAI-compatible)
Web search	Tavily API

Getting started

Choose the setup that matches your workflow.

---

Option A — Docker (recommended, zero local dependencies)

Starts Postgres + Ollama in containers and runs the app natively.

# 1. Start infrastructure
docker compose up -d

# 2. Install JS deps and sync the schema
pnpm install
pnpm db:push

# 3. Pull an AI model (first time only — ~2 GB download)
docker compose exec ollama ollama pull llama3.2

# 4. Copy env and start the app
cp apps/web/.env.example apps/web/.env.local
pnpm dev

apps/web/.env.local already has the right defaults for the Docker services. Open http://localhost:3000.

---

Option B — Native (everything runs on your machine)

Prerequisites

• Node.js 20+
• pnpm (npm install -g pnpm)
• PostgreSQL 14+ running locally
• Ollama installed and running

Steps

# 1. Clone and install
git clone <repo>
cd bitebase
pnpm install

# 2. Configure environment
cp apps/web/.env.example apps/web/.env.local

Edit apps/web/.env.local:

DATABASE_URL=postgresql://postgres:postgres@localhost:5432/bitebase

# Generate with: openssl rand -hex 32
BETTER_AUTH_SECRET=your-32-char-secret-here
BETTER_AUTH_URL=http://localhost:3000
BETTER_AUTH_TRUSTED_ORIGINS=http://localhost:3000

OLLAMA_BASE_URL=http://localhost:11434/v1
OLLAMA_MODEL=llama3.2

# Optional — enables web search during lesson generation
TAVILY_API_KEY=tvly-your-key-here

# 3. Create database and sync schema
createdb bitebase
pnpm db:push

# 4. Pull a model and start Ollama
ollama pull llama3.2
ollama serve

# 5. Start the web app
pnpm dev

Open http://localhost:3000.

---

Mobile app (optional)

cp apps/mobile/.env.example apps/mobile/.env.local
# Set EXPO_PUBLIC_API_URL=http://<your-local-ip>:3000

pnpm --filter @bitebase/mobile start

Scan the QR code with Expo Go.

---

Full Docker stack (production-like)

Builds and runs everything in containers — useful for staging or smoke-testing the production build:

export BETTER_AUTH_SECRET=$(openssl rand -hex 32)
docker compose -f docker-compose.prod.yml up --build

# First time — pull the model
docker compose -f docker-compose.prod.yml exec ollama ollama pull llama3.2

Open http://localhost:3000.

Project structure

bitebase/
├── apps/
│   ├── web/              # Next.js 15 web app
│   └── mobile/           # Expo React Native app
├── packages/
│   ├── ai/               # Vercel AI SDK config, prompts, Zod schemas
│   ├── api/              # tRPC router + Better Auth
│   ├── db/               # Drizzle ORM + PostgreSQL schema
│   └── ui/               # Shared UI components
└── turbo.json

How it works

1. Onboarding — Chat with BiteBase. It asks about your topic, experience level, goals, and available time.
2. Curriculum generation — The AI generates a structured lesson plan with sections, subsections, and automatically classifies each course into a broad category and a more specific subcategory (owner‑editable).
3. Content generation — For each subsection, the AI optionally searches the web (Tavily), then writes a markdown lesson and a quiz.
4. Learning — Work through lessons in order. Each lesson is unlocked after passing the previous quiz (≥70%).
5. Progress — Your scores, streaks, and completion are tracked per user.

Swapping models

Set OLLAMA_MODEL in .env.local to any model Ollama supports:

OLLAMA_MODEL=mistral
OLLAMA_MODEL=gemma3
OLLAMA_MODEL=deepseek-r1

Any OpenAI-compatible API works too — just set OLLAMA_BASE_URL and OLLAMA_MODEL accordingly.