---
title: 'AI Flashcard Generator '
sdk: streamlit
emoji: 📚
colorFrom: blue
colorTo: purple
pinned: false
sdk_version: 1.58.0
app_file: app.py
---
# AI-Powered Flashcard Generator

A Streamlit web app that turns lecture notes into exam-revision flashcards.

## Features

- Upload lecture notes as PDF, TXT, or image files.
- Extract and clean raw text with PyMuPDF for PDFs.
- Split notes into 300-500 token chunks.
- Summarise chunks with a Hugging Face T5/BART model.
- Use LangChain prompt templates for a three-step generation pipeline:
  - Extract key concepts as JSON.
  - Generate Q&A flashcards as JSON.
  - Optionally rewrite easy cards into harder Bloom's Taxonomy questions.
- Export flashcards as Anki-compatible CSV.
- Optionally export `.apkg` when `genanki` is installed.

## Setup

```bash
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
```

## Run

```bash
streamlit run app.py
```

The first summarisation run can take a while because Transformers may download the selected model.

## Test

```bash
pytest
```

## Notes

- Image extraction uses `pytesseract` when it is installed and Tesseract OCR is available on your system.
- If Hugging Face model loading fails, the app falls back to a deterministic extractive summary so the project remains usable offline.
- CSV exports can be imported into Anki with the first field as `Front` and the second field as `Back`.