Duos is a Next.js-based educational platform designed to digitalize traditional German language textbooks. It transforms static PDFs into interactive, self-correcting digital workbooks using a multi-stage LLM processing pipeline.
The core of this project is a three-stage transformation process:
- Extraction: Raw Textbook PDFs are converted into clean, structural Markdown using specialized extraction prompts (
prompt1.md). - Answer Alignment: Answer keys are extracted and normalized into a consistent Markdown format (
prompt1.md). - Structuring: An LLM-orchestrated prompt (
prompt2.md) merges the content and answers into a typed json file. - Deployment: The JSON is served by the Next.js frontend to render interactive Fill-in-the-blank, Multiple Choice, and Vocabulary modules.
- Zero-Hustle Answers: Answers are automatically mapped to exercises; no more flipping to the back of the book.
- Interactive Exercises: Self-correcting inputs for German grammar and vocabulary.
- Structured Learning: Vocabulary is categorized by gender (der/die/das) and part of speech automatically.
- Markdown Mid-layer: Allows for easy auditing of content before it enters the application database.
/public/textbooks/: Stores the final processed JSON files./prompts/: Contains the LLM engineering logic./components/: React components designed to handle different question types (Matching, Fill-blank, etc).
- Clone the repo:
git clone https://github.com/orionbits/duos/ - Install dependencies:
npm install - Run the development server:
npm run dev
The project aims to reduce the time it takes to "digitize" one chapter from hours of manual entry to minutes of automated AI processing.