Caelune — AI Study Assistant

Transform lecture notes into exam-ready summaries using free Hugging Face inference.

Features

• 3 output formats — Bullet Points, Flashcards Q&A, Structured Study Guide
• File uploads — PDF, Word (.docx), PowerPoint (.pptx), plain text, images (PNG/JPG)
• Image OCR — Upload a photo of handwritten notes or a textbook page; text is extracted automatically
• Key terms extractor — Auto-generates a glossary after every summary
• 12 subjects — General, Biology, Chemistry, Physics, Maths, CS, Medicine, Law, Economics, Psychology, History, Literature
• 3 difficulty levels — Beginner, Intermediate, Expert
• 3 response lengths — Concise, Standard, Detailed
• 10 languages — English, Spanish, French, German, Hindi, Portuguese, Arabic, Japanese, Korean, Mandarin
• Session history — Browse and restore previous summaries
• Download — Save summaries and key terms as plain text

Quick Start (local)

git clone https://github.com/YOUR_USERNAME/caelune.git
cd caelune
pip install -r requirements.txt

Create .streamlit/secrets.toml:

HF_TOKEN = "hf_your_token_here"

Get a free token at huggingface.co/settings/tokens.

streamlit run streamlit_app.py

Deploy to Streamlit Community Cloud

1. Push this repo to GitHub (make sure .streamlit/secrets.toml is in .gitignore)
2. Go to share.streamlit.io → New app → select this repo
3. In Secrets, add: HF_TOKEN = "hf_your_token_here"
4. Deploy

Tech Stack

Layer	Technology
UI	Streamlit 1.35+
LLM	Hugging Face Inference API (free serverless tier)
Default model	Qwen 2.5 7B Instruct
Vision/OCR	Llama 3.2 11B Vision Instruct
PDF parsing	pdfplumber
Word parsing	python-docx
PPT parsing	python-pptx
Markdown rendering	Markdown (Python)

Models

Name	Use
Qwen 2.5 7B Instruct	Default summarisation
Zephyr 7B Beta	Alternative
Phi-3.5 Mini Instruct	Lightweight
Gemma 2 2B Instruct	Lightweight
Llama 3.2 11B Vision	Image OCR (auto-selected)

License

Apache 2.0