RAG guide
Graph RAG compression
Graph RAG retrieves information by traversing entity relationships in a knowledge graph. The traversed path can include many nodes, each adding context. Compression keeps the relevant nodes and drops the rest.
Graph traversal context
A graph RAG query might traverse: User → Account → Transactions → Merchant → Reviews. Each node's description adds 100-500 tokens. A traversal with 10 nodes adds 1,000-5,000 tokens. Only the nodes most relevant to the query need to be in the LLM context.
Frequently asked questions
Does compression preserve graph relationships?
Yes. The relationships between preserved nodes are maintained in the compressed output.
Can I use it with Neo4j or Amazon Neptune?
Yes. Export the graph traversal results as text, compress, then send to the LLM.
Try it yourself
Paste your long prompt into the playground, ask a question, and see what SuperCompress keeps and removes. Free, no signup needed.