The Problem :-. During exam time and high stress periods , students find it very difficult to read a masssive book in a short amount of time .

Our Solution

Our AI book to quiz generator is an end-to-end pipeline and web app that takes a PDF book and produces a ready-to-use ChromaDB vector store. It uses Hugging Face embeddings, can be tested in the browser, and exports as a complete, runnable VS Code project.We built a semantic search or RAG (Retrieval-Augmented Generation) system using boilerplate: it used technologies like extracting text, splitting it into sensible chunks, choosing an embedding model, setting up a vector database, written all together.

Key Features
PDF ingestion: Upload a book and extract its text page by page, plus embedded images, with page numbers preserved for source citations.
Smart chunking: Text is split into chunks suited to embedding, so each vector represents a coherent piece of the book.
Hugging Face embedding pipeline: Chunks are converted into dense vectors using open-source models, with no vendor lock-in.
ChromaDB vector store generation: Embeddings are stored in a persistent Chroma collection, ready for similarity search.
In-browser interactive testing: Ask questions against your book right in the UI and see the retrieved chunks, so you can check quality before exporting.
One-click VS Code project export: The whole pipeline is bundled into a ZIP (via JSZip) containing code, config, and instructions. Unzip it, open it in VS Code, and run it.
Gemini integration (server-side): The Gemini API is used on the server to power AI-assisted features without exposing API keys to the client.
How It Works
Upload: The user drops a PDF into the React frontend.
Extract: The file goes to the FastAPI backend (/api/process-pdf), which stores it and runs extract_pdf.py. This uses PyMuPDF to pull out per-page text and images, saved as structured JSON plus an assets folder.
Chunk and embed: The text is split into chunks, and each chunk is embedded with a Hugging Face model.
Store: Vectors and metadata (source, page, chunk ID) are written into ChromaDB.
Query and test: The user searches the book semantically in the browser and inspects the top-k results.
Export: The generated pipeline is packaged as a downloadable VS Code project for local use or further development.