🤖 Research Assistant with RAGs

An intelligent assistant that helps you explore academic papers from arXiv using Retrieval-Augmented Generation (RAG).

Overview

🔍 Features

• 🔽 Automatic paper download from arXiv by topic or arXiv ID (2046 files)
• 🧠 Embedding generation using OpenAI’s text-embedding-3-small model
• 🗃️ Storage in Qdrant, a high-performance vector database
• 💬 Question answering powered by GPT-4 with context retrieved via RAG
• 📚 Multi-paper search and chunk-level semantic retrieval

Demo

Known Issues

• The system does not save the prompt history, which causes the assistant’s responses to lack context and sometimes become confusing.

• The PDF-to-text conversion and text chunking processes are not yet optimized, resulting in inconsistent context that makes the assistant’s replies appear incoherent or “hallucinated.”

• The number of paper files available is still too limited, restricting the assistant’s knowledge base and accuracy.