I decided to build this chatbot, with the help of Real Python's LLM RAG Chatbot tutorial, to have an LLM project to build upon as I learn new topics and experiment with new ideas.
Along the way, I learned about LangChain, how and when to use knowledge graphs, and how to quickly deploy LLM RAG apps with FastAPI and Streamlit.
Start by cloning this repo and adding a .env file with the following environment variables:
OPENAI_API_KEY=<YOUR_OPENAI_API_KEY>
NEO4J_URI=<YOUR_NEO4J_URI>
NEO4J_USERNAME=<YOUR_NEO4J_USERNAME>
NEO4J_PASSWORD=<YOUR_NEO4J_PASSWORD>
HOSPITALS_CSV_PATH=https://raw.githubusercontent.com/hfhoffman1144/langchain_neo4j_rag_app/main/data/hospitals.csv
PAYERS_CSV_PATH=https://raw.githubusercontent.com/hfhoffman1144/langchain_neo4j_rag_app/main/data/payers.csv
PHYSICIANS_CSV_PATH=https://raw.githubusercontent.com/hfhoffman1144/langchain_neo4j_rag_app/main/data/physicians.csv
PATIENTS_CSV_PATH=https://raw.githubusercontent.com/hfhoffman1144/langchain_neo4j_rag_app/main/data/patients.csv
VISITS_CSV_PATH=https://raw.githubusercontent.com/hfhoffman1144/langchain_neo4j_rag_app/main/data/visits.csv
REVIEWS_CSV_PATH=https://raw.githubusercontent.com/hfhoffman1144/langchain_neo4j_rag_app/main/data/reviews.csv
HOSPITAL_AGENT_MODEL=gpt-3.5-turbo-1106
HOSPITAL_CYPHER_MODEL=gpt-3.5-turbo-1106
HOSPITAL_QA_MODEL=gpt-3.5-turbo-0125
CHATBOT_URL=http://host.docker.internal:8000/hospital-rag-agent
Next, navigate to the project root, start Docker, make sure your AuraDB instance is up and running, and then run:
make build
The build will start the servers, however, you can also start the server with the command:
make start
and stop all containers by running:
make stop
You can interact with the chatbot API docs on localhost:8000/docs, and the UI on localhost:8501:
It would be useful to look at the way the database has been structured to come up with some potential questions you want to make to the chatbot:
