PDF to Lesson.

Upload a PDF and instantly generate a small interactive course with modules, lessons, quizzes, and flow-ordering questions. Powered by Together AI.

How it works

PDF to Interactive Lesson Generator extracts text from your uploaded PDF, sends the source content to Together AI-hosted models, and builds a structured 3-module course. Each module includes grounded lesson content plus interactive checks like short-answer, true-false, multiple-choice, and process-ordering questions.

1. Upload a PDF from the browser
2. Extract — the app reads the document text locally with MuPDF
3. Generate — Together AI models create the course structure, lessons, quizzes, and flow questions
4. Repair — duplicate or weak questions are detected and regenerated or hidden
5. Learn & Share — courses are saved so you can revisit progress or share a public course link

Tech stack

• Framework: Next.js App Router
• AI: Together AI — course planning, lesson generation, quiz generation, and flow-question generation (GPT OSS 120B)
• PDF extraction: MuPDF
• Database: Neon Postgres with Drizzle ORM
• Storage: Vercel Blob
• Queueing and rate limiting: Upstash Redis
• Styling: Tailwind CSS
• UI: Radix UI, lucide-react, and custom interactive lesson components

Under the hood

The generation pipeline is designed to be fast enough for a live demo while keeping the generated course grounded in the source PDF.

• OCR and extraction: MuPDF pulls text from the uploaded PDF before generation starts
• Course planning: one AI call creates the module structure
• Flow assignment: one AI call finds distinct source-backed processes for flow-ordering questions
• Parallel lesson generation: modules are generated concurrently to reduce wait time
• Dedup repair: similar questions are detected with Jaccard similarity, then regenerated or removed from the visible course
• Persistence: generated courses are stored in Postgres and PDF uploads are handled through Vercel Blob

For a deeper breakdown of the speed and quality work, see docs/course-generation-speedup.md.

Running Locally

Cloning the repository

git clone https://github.com/Nutlope/pdf-to-interactive-lesson.git
cd pdf-to-interactive-lesson

Getting API keys

Together AI (required — powers course generation):

1. Go to Together AI to create an account
2. Copy your API key

Neon (required — stores generated courses):

1. Create a Postgres database with Neon
2. Copy your database connection string

Vercel Blob (required for the web app — stores uploaded PDFs):

1. Create or connect a Vercel Blob store
2. Copy your read/write token

Upstash Redis (required for queueing and rate limiting):

1. Create an Upstash Redis database
2. Copy the REST URL and REST token

Storing API keys in .env

Create a .env.local file in the root directory and add your keys:

TOGETHER_API_KEY=your_together_api_key_here
DATABASE_URL=your_neon_database_url_here
BLOB_READ_WRITE_TOKEN=your_vercel_blob_token_here
UPSTASH_REDIS_REST_URL=your_upstash_redis_rest_url_here
UPSTASH_REDIS_REST_TOKEN=your_upstash_redis_rest_token_here
NEXT_PUBLIC_APP_URL=http://localhost:3000

You can also enter a Together AI key directly in the app by clicking the key icon in the top-right corner.

Installing dependencies

pnpm install

Setting up the database

pnpm db:push

Running the application

pnpm dev

The app will be available at http://localhost:3000.

CLI

This repo also includes a small CLI for local generation, debugging, and benchmarking.

pnpm course generate data/document.pdf
pnpm course modules data/document.pdf
pnpm course benchmark data/document.md --runs 5

Useful flags:

• --model <name>
• --output <path>
• --save-text-auto
• --no-validate
• --no-retry
• --max-retries <n>
• --verbose