# HER2 Q&A Chatbot

This chatbot allows you to ask questions about a scientific paper (HER2 breast cancer study). It uses:

- FAISS vector search with sentence-transformer embeddings
- RAG with local LLM (TinyLlama) for Q&A
- Streamlit for the UI
- Reranking with BAAI bge-reranker-base

## Configuration

```yaml
sdk: streamlit
app_file: app.py
runtime: python
```

## How it works

1. The PDF is parsed and split into chunks.
2. Chunks are embedded and stored in FAISS.
3. A user query retrieves relevant chunks.
4. Optionally, a reranker improves the chunk order.
5. The top chunks and query are used to generate a response with TinyLlama.

## Running locally

```bash
pip install -r requirements.txt
streamlit run app.py
```

## Hosted Demo

👉 Available on Hugging Face Spaces!