RAG Web App Project

This project focuses on developing a web application that enables private corporations to retrieve documentation. The application allows users to obtain information about variables or techniques used in their technologies through question answering, utilizing Large Language Models (LLMs).

This research was conducted as part of the EU Horizon project SEUS – Smart European Shipbuilding (Grant Agreement No. 101096224), funded by the European Union.

Getting Started

Follow these steps to set up the project locally:

1. Clone the repository (You can use HTTPS too):

   git clone [email protected]:TurkuNLP/RAG-web-app.git
   

2. Create a virtual environment :

   python3 -m venv env
   

Note: Replace env with your preferred environment name.

3. Activate the virtual environment :

   source env/bin/activate
   
   

4. Install dependencies :

   pip install -r /path/to/requirements.txt
   

Run an App (Single Entry Point)

Use run.py to select which Flask app to run with APP_NAME. Optional PORT, HOST, and DEBUG environment variables control runtime.

Supported apps (APP_NAME):

• local
• seus
• arch-ru
• arch-en
• news
• law

Examples:

APP_NAME=local PORT=5000 python run.py
APP_NAME=law PORT=8080 python run.py
APP_NAME=arch-en python run.py

Docker

Build the image:

docker build -t rag-web-app .

Run a selected app:

docker run --rm \
  --env-file .env \
  -v "$PWD:/app" \
  -w /app \
  rag-web-app \
  python python_script/populate_database.py --config local

docker run --rm \
  --env-file .env \
  -v "$PWD/data:/app/data" \
  -p 8000:8000 \
  -e APP_NAME=local \
  -e PORT=8000 \
  rag-web-app

Using Docker Compose:

docker compose up --build

Change the app by editing APP_NAME in docker-compose.yml or passing -e APP_NAME=law to docker run.

Conda Environment (Optional)

If you use Conda, create an environment and install Python dependencies with pip:

conda create -n rag-web-app python=3.11
conda activate rag-web-app
pip install -r requirements.txt

Configuration (config.json)

config.json stores configuration settings for different environments or use cases. Each configuration specifies:

• data_path: Path to the data folder.
• chroma_root_path: Path where the Chroma database will be stored.
• embedding_model: Name of the model used for embeddings.
• llm_model: Name of the language model to be used.

Environment Variables

API keys are loaded from a .env file via python_script/parameters.py. Common keys:

• OPENAI_API_KEY: required when using OpenAI models.
• HF_API_TOKEN: required when using Hugging Face hosted models.
• VOYAGE_API_KEY: required when using VoyageAI models (if enabled).

Runtime settings:

• APP_NAME: which app to run (see Supported apps above).
• HOST: bind address (default 0.0.0.0).
• PORT: port to listen on (defaults to the app’s configured port).
• DEBUG: set to true/1 to enable Flask debug (local dev only).

Database Scripts (python_script/)

Four files in python_script/ are related to database setup and management:

• parameters.py: loads parameters from config.json.
• get_embedding_function.py: loads the embedding model.
• populate_database.py: create, reset, or clear the database.

populate_database.py

The main() function is the CLI entry point for database management.

Arguments:

• --config (str): configuration name in config.json.
• --reset: clear the database subfolder for the config, then repopulate.
• --clear: clear the database (optionally scoped to the config).

Functionality:

• Populate: --config only loads documents, splits them, and adds them to Chroma.
• Reset: --config --reset clears the config’s subfolder, then repopulates.
• Clear: --clear removes data; if --config is provided, it targets the config’s subfolder (based on EMBEDDING_MODEL).

Example usage:

python populate_database.py --config CONFIG_NAME
python populate_database.py --config CONFIG_NAME --reset
python populate_database.py --config CONFIG_NAME --clear