The dataset behind memdb-oracle

Sanity project ynsa3nyp, dataset production, public. Everything below is read live from the Sanity API with no token. ← back to the agent

One benchmarkRun document per measured run, with every knob that moved the number as a field, plus engine, finding and benchmarkSession documents. The agent answers numeric questions with GROQ over these; prose comes from a Sanity Context Knowledge Base built from the write-ups and the vendors' own pages.

Runs per engine

Engine	Version tested	Threading model	Runs	Notes

Headline findings (with the caveat the agent must repeat)

Finding	Caveat	Vendor claim it is compared with

What one run looks like

loading…

Query it yourself

Any GROQ works against the public endpoint. Example, the headline configuration:

Raw JSON for this page: overview query · Studio (members only): memdb-oracle.sanity.studio · Source CSVs and harness: GitHub