# RVBBIT

> Open-source Postgres extension: semantic operators in plain SQL (means, summarize, extract, triples), table time travel, columnar acceleration with adaptive engine routing, and receipts for every model-backed answer. With a Clover key, per-row inference runs on hosted GPUs under a flat subscription — agents orchestrate in SQL without spending their own tokens.

Everything is reachable over a normal Postgres connection; discovery is SQL
(`rvbbit.operators`, `rvbbit.capability_catalog`, `rvbbit.capability_search(text)`).

## Setup

- [Setup for Agents](https://rvbbit.ai/agents.md): install, get a key, verify, example queries — written for agents
- [Clover install SQL](https://rvbbit.ai/clover-install.sql): idempotent psql script registering the hosted operators
- [Quickstart](https://rvbbit.ai/docs/quickstart): the Docker ensemble and first semantic queries
- [install.sh](https://rvbbit.ai/install.sh): full-stack one-liner (Postgres 18 + RVBBIT + Data Rabbit desktop)

## Reference

- [Capability catalog (JSON)](https://rvbbit.ai/catalog.json): machine-readable list of operators, models, and install statements
- [Clover](https://rvbbit.ai/clover): hosted operator sets and industry kits — pricing, install, design partners
- [Docs index](https://rvbbit.ai/docs): human documentation
- [Benchmarks](https://rvbbit.ai/benchmarks): ClickBench / TPC-H route-by-route results

## Optional

- [Data Rabbit](https://datarabbit.com): the SQL desktop for the humans you work with