Comparison to multi-hop-qa with base AutoSchemaKG approach

[adsharma]

Looks like your original approach was to provide graph database query based context to a LLM to improve multi-hop query performance.

How does that approach compare to AtlasKV? Do you have any benchmark data on performance or price/performance? I can imagine that the latency could be lower with AtlasKV.