memdb-oracle

by Sunny Sahijwani, Two Techies · an agent over an independent in-memory database benchmark

Answers about Redis, Valkey, Dragonfly, KeyDB and Garnet from 478 measured benchmark runs (Sanity dataset, GROQ) and a Knowledge Base that shows vendor claims next to measurements. Every answer ends with the GROQ it ran. A turn takes 30–60 s: it is several Sanity Context calls plus reasoning; tool activity streams while you wait.

Numbers come only from GROQ over 478 measured runs in a public Sanity dataset (project ynsa3nyp, browse it, no login needed); explanations and vendor claims only from a Sanity Context Knowledge Base, cited per sentence. Anything explained from general knowledge is labelled as such.
Read the benchmark series · Rerun the benchmark · Agent source (MIT) · Built for the Sanity Challenge with Sanity Context and the Claude API.