FUNDus! MuRAG

This repo contains the code for the FUNUus! MuRAG System -- a multimodal RAG-based Assistant to explore the FUNDus! database.

Starting the system in Production Mode

Prerequisites

Getting the data

Note that this only works for LT and HCDS members. If you are not a member, you can request access to the data by contacting the Florian Schneider.

1. Create a folder data in the root of the project
2. Copy the DataFrames from /ltstorage/shares/projects/fundus-murag/data to the data folder
3. Contact Florian Schneider to get the Google Service Acoount Credentials file and place it in the data folder (or create a new one with your Google Cloud account)

Setting up the Production environment

1. Navigate to the docker folder
2. Run ./setup-folders.sh to create the necessary folders for the Docker volumes
3. Edit the .env.example file and save it as .env with the modified values

Starting the Docker containers

1. Run docker compose pull to pull the necessary images
2. Run docker compose up -d to start the system
3. Run docker compose logs -f to see the logs of the running containers

Starting the system in Development Mode

Prerequisites

1. Create a virtual environment with Python 3.10
2. Navigate to repository root
3. Install the requirements with pip install -r requirements.txt
4. Get the data as described in the Docker section
5. Navigate to the docker folder
6. Run ./setup-folders.sh to create the necessary folders for the Docker volumes
7. Edit the .env.dev.example file and save it as .env.dev with the modified values
8. Navigate back to the repository root
9. Edit the config/config.dev.example.yaml file and save it as config/config.dev.yaml with the modified values

(OPTIONAL) Starting the ML Service in Development Mode

You only need to do this if you want to run the ML service locally, e.g., to change the Embedding Model or add new functionality. Otherwise, skip this.

1. Navigate to repository root
2. In the src/fundus_murag/ml/server.py file, set the port to a random number (e.g., 23456)
3. Run CUDA_VISIBLE_DEVICES=1 PYTHONPATH=src FUNDUS_ML_DEV_MODE=1 python src/fundus_murag/ml/server.py to start the ML service in development mode
4. Check the logs to see on which port the service is running. You will need this port to start the MESOP application

Starting the Docker Containers for Development

1. Navigate to the docker folder
2. (Optional) If you run the ML service in development mode, remove the fundusml profile from the .env.dev
3. Run docker compose --env-file .env.dev up in a tmux or similar shell to start the dev containers (i.e., Weaviate)
4. Run curl http://localhost:<YOUR_FUNDUS_ML_EXPOSED_PORT>/embed to check whether the FUNDus ML Service is up and running. This should print sth. like {"detail":"Method Not Allowed"}

Starting the FUNDus MuRAG Application for Development

1. Navigate to repository root
2. Select a port of your choice (e.g., the FUNDUS_UI_EXPOSED defined in the docker/.env file)
3. Run FUNDUS_CONFIG_FILE=config.dev.yaml PYTHONPATH=src mesop --port <PORT OF YOUR CHOICE> src/fundus_murag/ui/main.py to start the FUNDus MuRAG application.
4. Open your browser and navigate to http://localhost:<PORT OF YOUR CHOICE> to see the application. Don't forget to forward the port if you are running the application on a remote server.