MünsterGPT

project image

Overview

As part of 'Münsterhacks 23', I was allowed to develop a chatbot for a small group that can answer questions about the city of Münster's citizen service. To do this, we first crawled the relevant websites of the city of Münster and converted this information into vector embeddings using the OpenAI interface and then stored these vectors in a ChromaDB database. A GPT-4 model was then enriched with this context in order to obtain the necessary knowledge. We then provided a hosted chatbot via a simple graphical interface with the help of LangChain and Streamlit, with which we were even able to win second prize in the hackathon. The project is also being funded by the city of Münster and is currently being further developed.