An LLM-based system that answers questions about a corpus of research papers from the arXiv.
In this project, I build a system based on OpenAI's GPT models to answer questions over a corpus of documents. For illustration I use papers downloaded from the arxiv at https://arxiv.org and extract content using langchain's DocumentLoader, on Lithium Ion Batteries, but you can easily modify this for any content --- journal entires, research notes, movie scripts, anything really.
The easiest way to interact with the codebase is to switch to /src/ and then build the docker image with docker build -t qa-research-assistant ., and finally run the built container with docker run --rm qa-research-assistant -t TASK -f FIELD -o ORG_ID. For a full list of arguments, use docker run --rm qa-research-assistant -h.