Running RAG Me Up

To run RAG Me Up you need Postgres (with pgvector + ParadeDB pg_search), the Python RAG backend, and a client. The recommended path is Docker Compose, which starts ParadeDB, the Python server, the Node.js API, and the React UI together. For GPU embeddings/inference, use hybrid mode so the Python server runs on the host. Details and env templates live in the project README.

Docker Compose (CPU)

cp docker-compose.env.example docker-compose.env
# Set at least POSTGRES_PASSWORD and JWT_SECRET

docker compose --env-file docker-compose.env up --build -d

The React UI is then available on http://localhost (or your configured HOST_PORT).

Hybrid mode (GPU for Python)

Postgres, Node, and React stay in Docker; the Python RAG server runs on the host with CUDA access:

docker compose --env-file docker-compose.env \
  -f docker-compose.yml -f docker-compose.hybrid.yml up --build -d

cd server
python -m venv .venv
# activate the venv, then:
pip install -r requirements.txt
# ensure server/.env has postgres_uri pointing at localhost:6024 and embedding_cpu=False
python server.py

Running the Python backend alone

Useful when you already have Postgres and only want the RAG API:

1. Create a virtual environment using venv or similar.
2. Navigate to the /server folder in a shell.
3. Install the requirements through pip install -r requirements.txt.
4. Configure .env (see Configuring the .env File).
5. python server.py to start the server.

Running the React / Node UI alone

1. Navigate to /ui/node-react.
2. npm run install:all (installs both server and client dependencies).
3. Start the Node API (npm run dev:server or npm run start:server) and the React client (npm run dev:client).
4. Point the Node layer at your Python server URL as documented in the Compose / hybrid env files.

Make sure Postgres is up before the Python backend, and that the Python backend is reachable before you expect the UI to answer questions. The Python server is stateless; conversation memory lives in the client.