Meal Planner GraphRAG

Overview

This repository showcases the implementation of a GraphRAG (Graph-based Retriever-Augmented Generation) meal planning assistant. The project leverages a Neo4j graph database, semantic search, and LLM-driven reasoning to generate personalized recipe recommendations and shopping lists based on user preferences and dietary needs.

Architecture

Getting Started

Prerequisites

• Docker & Docker Compose
• Python 3.10+
• OpenAI API key (for LLM features)
• Neo4j (runs in Docker)

Setup

1. Clone the repository:

   git clone <repo-url>
   cd meal-planner-graphrag

2. Configure environment variables:

• Example environment files are provided as .env.example and .env-docker.example in the app/ directory, and .env.neo4j.example in the neo4j/ directory.
• To configure your environment:
  1. Copy the example files to their active counterparts:

  cp app/.env.example app/.env
  cp app/.env-docker.example app/.env-docker
  cp neo4j/.env-docker.example neo4j/.env-docker

  2. Edit these files to set your API keys, database credentials, Neo4j settings, and other configuration as needed.
• These files are required for both local development and Docker-based deployment.

3. Start Neo4j and the app:

   docker-compose up --build

4. Access the application:

   • The app will be available at http://localhost:8501
   • The neo4j database will be available at http://localhost:7474

Development

• Main application logic is in app/app.py.
• Prompts and LLM logic are in app/core/prompts.py.
• Evaluation tools are in evaluation/.

Evaluation

• Use evaluation/eval.ipynb to assess the quality of generated research plans and retrieval outputs.