This project is a Retrieval-Augmented Generation (RAG) Chatbot for Scientific Papers designed to assist users in navigating, summarizing, and understanding scientific literature. Built as a web app, this tool allows users to ask complex questions about scientific content, receive precise answers, and engage with large volumes of research material more interactively and efficiently.
The RAG system was evaluated with the help of the deepeval Python package. Note that the evaluation module requires additional dependencies.
- Install dependencies:
pip install -r requirements.txt - Install Ollama - https://ollama.com/download
- Download LLM -
ollama pull qwen2.5:7b-instruct-q4_0 - Run app:
python -m ChatBot/streamlit run app.py
- Frontend: Streamlit (Python-based web app framework for simplicity and speed).
- Backend: LangChain for modular integration with LLMs and retrieval components.
- LLM Serving:: Ollama, enabling efficient model deployment and performance.
We’ve included a Python script, pdf2faiss.py, which allows users to create their own vector stores and customize the chatbot to suit their specific needs. Once the vector store is created, simply update its path in app.py on line 17.