Answer from a small part of the graph, not the whole graph.
A question names genes, diseases or drugs. The pipeline keeps only the part of a biomedical knowledge graph around them and has a language model answer from summaries of that part. On the 100 BioASQ questions of the evaluation, counting detection, indexing and search, the median question used about 7 times fewer tokens than GraphRAG on the whole graph when the subgraph has only the neighbours of those entities, and 20% fewer with the full subgraph, which spent more on 43 of the 100.
Type a question, or replay one already answered.
Pick a question on the left to watch how it was answered.
Each step of the pipeline appears here as it happens: the entities found in the question, the subgraph grown around them, the community reports written over it, and the final answer with its sources.