Query Benchmark Comparison

[zard1152]

From the benchmark results, I observed that this project both of insertion and query quality is better than LightRAG and GraphRAG. However, both LightRAG and GraphRAG have a common issue: their query time is too long. This has been a major concern for me. I would like to see a comparison chart of query time costs for better clarity.

[liukidar]

Hello, the query procedure is very similar, so the times should be identical. The context size can be reduced (maybe only providing chunks for example) to ensure a faster generation of the answer.
The latency can also be reduced by using local language models (in particular since we "read" the query we need to wait for one LLM call)

[Authorlove]

Is the processing time of PPR (Personalized PageRank) substantial when dealing with large graphs?