intersystems-iris-graphrag

This is a basic template (created during the 2025 InterSystems Employee Hackathon), in order to run Graph-Rag natively in IRIS using it's Vector Search capabilities. It generates a working image for Jupyter + InterSystems IRIS. The image comes with all of the necessary python modules installed to use graphrag out of the box.

Description

This repository provides a ready-to-go development environment for using our graphrag implementation.

Prerequisites

Make sure you have git and Docker desktop installed.

Installation

1. Clone the Repository

git clone https://github.com/fanji-isc/IRIS-GraphRAG.git
cd IRIS-GraphRAG

2. Set up the Environment

Create a .env file inside the workspace/flask/ directory with the following contents:

# .env file
OPENAI_API_KEY=your-openai-api-key
SECRET_KEY=your-secret-key

You can generate a secure SECRET_KEY using:

python3 -c "import secrets; print(secrets.token_hex(32)"

Open the terminal in this directory and call the command to build and run the two containes: Note: Users running containers on a Linux CLI, should use "docker compose" instead of "docker-compose" See Install the Compose plugin

$ docker compose up -d

The UI should be accessible through http://127.0.0.1:5001/

The IRIS management portal should be accessible through: http://localhost:9092/csp/sys/UtilHome.csp

The IRIS code that we are using is in the src/GraphKB folder