Prerequisites
To begin, you’ll need to set up and familiarize yourself with Memgraph, the foundation of your GraphRAG system.
Getting started with Memgraph
To begin your GraphRAG journey, you’ll need Memgraph MAGE set up and ready to go. If you’re new to Memgraph, follow the getting started guide for detailed installation and setup instructions.
Requirements for GraphRAG
GraphRAG systems involve combining graph databases with LLMs. Here’s what you’ll need to set up for success:
- A running Memgraph MAGE instance
- Knowledge graph stored in Memgraph - identify your structured or unstructured data to convert into a knowledge graph and follow Memgraph’s data modeling docs to design your graph.
- LLM API - use the large language model of your choice.
Building a GraphRAG system involves multiple building blocks, and selecting the right ones isn’t straightforward.
Before you begin, carefully consider what information the LLM needs that it wouldn’t have without the knowledge graph. Identify the key context to provide, and prioritize the most relevant data for inference.
Remember, creating an ideal system rarely happens on the first try—iteration and evaluation are essential.
Hardware requirements
Since Memgraph is an in-memory graph database, the hardware requirements depend on the size of your dataset and the complexity of your queries.
- Memory: Sufficient RAM is essential for storing and processing your dataset in memory. To estimate your needs, refer to our RAM calculation guide.
- Storage: While Memgraph operates in memory, storage is required for backups, snapshots, and write-ahead logs (WAL) to ensure data durability. Learn more about best practices in our storage FAQ.
Next steps
Once your environment is ready:
- Proceed to Knowledge graph creation to model and import your data into Memgraph.
- Explore GraphRAG features to understand how to retrieve and manage knowledge effectively.
- Review Examples and demos for inspiration on building your own GraphRAG system.