A LLM/RAG application to chat with research papers
Still under development!
I downloaded a bunch of research papers (only abstract and metadata). From the abstracts, I compute embeddings using SentenceTransformers and store in a local Faiss vectorstore.
Currently, the frontend and backend have been seperated. The frontend is currently a simple Streamlit app, but I am working on making this into a nicer JS/React.
Backend is Python - using FastAPI with Langchain. The LLM used is API calls to OpenAI (gpt-3.5-turbo), but I am experimenting with using Ollama to run Llama2 or maybe Mistral models locally.