README_HF.md

metadata

title: PaperMate
emoji: 📄
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 5.49.1
app_file: app.py
pinned: false
license: mit

PaperMate - Your Personal Assistant for Exploring arXiv Papers

PaperMate helps you explore and understand research papers from arXiv using RAG (Retrieval-Augmented Generation) and advanced reranking.

Features

• 🔍 Semantic search across arXiv papers
• 🧠 LLM-powered question answering
• 📊 Advanced reranking for better results
• 💬 Interactive Gradio interface

How to Use

1. Enter your question about research papers in the text box
2. Click "Ask" to get an answer based on the arXiv knowledge base
3. The system will retrieve relevant papers, rerank them, and generate a comprehensive answer

Example Questions

• "What is the NEMO paper all about?"
• "Do you know something about Indian system to help Indians better understand each other?"
• "What are the latest developments in transformer architectures?"

Configuration

This Space uses the Hugging Face Inference API. You'll need to set the following secret:

• HUGGINGFACE_TOKEN: Your Hugging Face API token (get one at https://huggingface.co/settings/tokens)

The default model is mistralai/Mistral-7B-Instruct-v0.2, but you can change it in the settings.

Local Development

1. Clone the repository
2. Install dependencies: pip install -r requirements.txt
3. Download the vector store from here
4. Place the vector_store directory in the root
5. Set your HUGGINGFACE_TOKEN environment variable
6. Run: python app.py

Tech Stack

• Framework: LangChain
• UI: Gradio
• Vector Store: FAISS
• Embeddings: Sentence Transformers
• LLM: Hugging Face Inference API
• Reranking: Custom reranker with BM25

License

MIT